Social Distancing Correlations Analysis for San Jose

library(tidyverse)
library(plotly)
library(sf)
library(mapview)
library(tigris)
library(censusapi)
library(leaflet)
library(lehdr)


options(
  tigris_class = "sf",
  tigris_use_cache = TRUE
)

Sys.setenv(CENSUS_KEY="10dcd73d7c043e91bac9fb8d3989cbff54b08790")

Load social distancing data and blockgroups

Load the Safegraph social distancing data and San Jose blockgroups

# get SJ blockgroups 
# get San Jose block groups
scc_blockgroups <- block_groups("CA","Santa Clara", cb=F, progress_bar=F)

# Use tracts sent to us by San Jose
sj_tracts <- st_read("/Users/simonespeizer/pCloud Drive/Shared/SFBI/Data Library/San_Jose/CSJ_Census_Tracts/CSJ_Census_Tracts.shp") %>%
  st_as_sf() %>%
  st_transform(st_crs(scc_blockgroups))

## Reading layer `CSJ_Census_Tracts' from data source `/Users/simonespeizer/pCloud Drive/Shared/SFBI/Data Library/San_Jose/CSJ_Census_Tracts/CSJ_Census_Tracts.shp' using driver `ESRI Shapefile'
## Simple feature collection with 219 features and 9 fields
## geometry type:  MULTIPOLYGON
## dimension:      XY
## bbox:           xmin: 6112856 ymin: 1869687 xmax: 6255982 ymax: 1996555
## epsg (SRID):    2227
## proj4string:    +proj=lcc +lat_1=38.43333333333333 +lat_2=37.06666666666667 +lat_0=36.5 +lon_0=-120.5 +x_0=2000000.0001016 +y_0=500000.0001016001 +ellps=GRS80 +towgs84=0,0,0,0,0,0,0 +units=us-ft +no_defs

sj_citycouncil_disticts <- st_read("/Users/simonespeizer/pCloud Drive/Shared/SFBI/Data Library/San_Jose/City Council Districts/CITY_COUNCIL_DISTRICTS.shp") %>%
  st_as_sf() %>%
  st_transform(st_crs(scc_blockgroups))

## Reading layer `CITY_COUNCIL_DISTRICTS' from data source `/Users/simonespeizer/pCloud Drive/Shared/SFBI/Data Library/San_Jose/City Council Districts/CITY_COUNCIL_DISTRICTS.shp' using driver `ESRI Shapefile'
## Simple feature collection with 10 features and 7 fields
## geometry type:  POLYGON
## dimension:      XY
## bbox:           xmin: 6112856 ymin: 1869687 xmax: 6255982 ymax: 1996555
## epsg (SRID):    2227
## proj4string:    +proj=lcc +lat_1=38.43333333333333 +lat_2=37.06666666666667 +lat_0=36.5 +lon_0=-120.5 +x_0=2000000.0001016 +y_0=500000.0001016001 +ellps=GRS80 +towgs84=0,0,0,0,0,0,0 +units=us-ft +no_defs

# from code written by others to get SJ blockgroups
sj_blockgroups <-
  scc_blockgroups %>%
  st_centroid() %>%
  st_join(sj_tracts, left = F) %>%
  st_join(sj_citycouncil_disticts%>% dplyr::select(DISTRICTS)) %>%
  mutate(
    DISTRICTS = DISTRICTS %>% factor(levels = c("1","2","3","4","5","6","7","8","9","10"))
  ) %>%
  st_set_geometry(NULL) %>%
  left_join(scc_blockgroups%>% dplyr::select(GEOID), by = "GEOID") %>%
  st_as_sf() %>%
  dplyr::select(GEOID, DISTRICTS)

# the spatial join leaves off two blockgroups which are touching district 9. The following code assigns those to district 9
sj_blockgroups$DISTRICTS[is.na(sj_blockgroups$DISTRICTS)] <- 9

# code from others in the class to get social distancing data 
sj_socialdistancing <- readRDS("/Users/simonespeizer/pCloud Drive/Shared/SFBI/Restricted Data Library/Safegraph/covid19analysis/sj_socialdistancing.rds") %>% 
  mutate(date = date_range_start %>%  substr(1,10) %>% as.Date()) %>% 
  left_join(sj_blockgroups, by = c("origin_census_block_group" = "GEOID")) %>% 
  filter(!is.na(DISTRICTS))

# obtaining weekends
weekends <-
  sj_socialdistancing %>% 
  filter(!duplicated(date)) %>% 
  arrange(date) %>% 
  mutate(
    weekend = 
      ifelse(
        (date %>% as.numeric()) %% 7 %in% c(2,3),
        T,
        F
      )
  ) %>% 
  dplyr::select(date,weekend)

sj_socialdistancing <- 
  sj_socialdistancing %>% 
  left_join(weekends)

# date of the shelter in place order
shelter_start <- "2020-03-16" %>% as.Date()

# get average staying at home on weekdays in January and February
sj_pre_sd_at_home_average <- sj_socialdistancing %>% 
  filter(weekend == F) %>% 
  filter(date <  as.Date("2020-03-01")) %>%
  group_by(origin_census_block_group) %>% 
  summarize(completely_home_device_count = sum(completely_home_device_count), device_count = sum(device_count)) %>% 
  mutate(`% Completely at Home Pre Shelter` = (completely_home_device_count/device_count*100) %>% round(1), `% not completely at home pre shelter` = (100 - `% Completely at Home Pre Shelter`))

Obtaining demographic variables

Here I obtain various demographic data, including income (percent below 50% and 80% of area median income), vehicle ownership, age, English language ability, and occupants per room.

# obtain the saved census data 
setwd("~/Documents/2020 Spring Quarter/CEE 218Z")
acs_vars = readRDS("censusData2018_acs_acs5.rds")
setwd("~/Documents/2020 Spring Quarter/CEE 218Z/covid19")

# load in income data - code adapted from other students
sj_median_income_by_block <-
  getCensus(
    name = "acs/acs5",
    vintage = 2018,
    region = "block group:*", 
    regionin = "state:06+county:085",
    vars = "B19013_001E"
  ) %>%
  mutate(
    blockgroup =
      paste0(state,county,tract,block_group)
  ) %>% 
  select_if(!names(.) %in% c("GEO_ID","state","county","tract","block_group","NAME")) %>% 
  rename(
    Median_Income = B19013_001E 
  ) %>% 
  filter(!is.na(Median_Income)) %>% 
  left_join(sj_blockgroups, by = c("blockgroup" = "GEOID")) %>% #this code gives each blockgroup a district designation
  filter(
    !is.na(DISTRICTS)
  ) %>% 
  
  # this code joins our census data with the social distancing data, processed as shown below
  left_join(sj_socialdistancing %>%  
                          filter(weekend == F) %>% 
                          filter(date > shelter_start) %>%
                          group_by(origin_census_block_group) %>% 
                          summarize(
                                    completely_home_device_count = sum(completely_home_device_count),
                                    device_count = sum(device_count)) %>% 
                          mutate(`% Completely at Home` = (completely_home_device_count/device_count*100) %>% round(1), 
                                 `% not completely at home` = (100 - `% Completely at Home`)),
            by = c("blockgroup" = "origin_census_block_group")
  ) %>% 
  filter(
    !is.na(device_count)
  ) %>% 
  left_join(sj_pre_sd_at_home_average %>% dplyr::select(origin_census_block_group, `% Completely at Home Pre Shelter`, `% not completely at home pre shelter`), by = c("blockgroup" = "origin_census_block_group"))

sj_ami_by_block <-
  getCensus(
    name = "acs/acs5",
    vintage = 2018,
    region = "block group:*", 
    regionin = "state:06+county:085",
    vars = "group(B19001)"
  ) %>%
  mutate(
    blockgroup =
      paste0(state,county,tract,block_group)
  ) %>% 
  select_if(!names(.) %in% c("GEO_ID","state","county","tract","block_group","NAME")) %>% 
  dplyr::select(-c(contains("EA"),contains("MA"),contains("M"))) %>%
  group_by(blockgroup) %>% 
  summarize(
    Total = B19001_001E,
    `Under 75,000` = sum(B19001_002E, B19001_003E, B19001_004E, B19001_005E, B19001_006E, B19001_007E, B19001_008E, B19001_009E, B19001_010E, B19001_011E, B19001_012E),
    #sum(lapply(2:12, function(x) as.name(paste0("B19001_00",x,"E"))))
    `Under 100,000` = sum(B19001_002E, B19001_003E, B19001_004E, B19001_005E, B19001_006E, B19001_007E, B19001_008E, B19001_009E, B19001_010E, B19001_011E, B19001_012E, B19001_013E), 
    `Under 125,000` = sum(B19001_002E, B19001_003E, B19001_004E, B19001_005E, B19001_006E, B19001_007E, B19001_008E, B19001_009E, B19001_010E, B19001_011E, B19001_012E, B19001_013E, B19001_014E)
  ) %>% 
  mutate(
    `% under 75,000` = `Under 75,000` / Total * 100,
    `% over 75,000` = (100 - `% under 75,000`),
    `% under 100,000` = `Under 100,000` / Total * 100,
    `% over 100,000` = (100 - `% under 100,000`),
    `% under 125,000` = `Under 125,000` / Total * 100,
    `% over 125,000` = (100 - `% under 125,000`),
  ) %>% 
  left_join(sj_median_income_by_block %>% dplyr::select(-Median_Income)
  ) %>% 
  filter(!is.na(device_count))

# loading in language data - code adapted from other students
sj_lang_by_block <-
  getCensus(
    name = "acs/acs5",
    vintage = 2018,
    region = "block group:*", 
    regionin = "state:06+county:085",
    vars = "group(B16004)"
  )  %>% 
  mutate(
    blockgroup =
      paste0(state,county,tract,block_group)
  ) %>% 
  select_if(!names(.) %in% c("GEO_ID","state","county","tract","block_group","NAME")) %>% 
  select(-c(contains("EA"),contains("MA"),contains("M"))) %>% 
  gather(
    key = "variable",
    value = "estimate", 
    - blockgroup
  ) %>% 
  left_join(acs_vars, by = c("variable" = "name")) %>% 
  mutate(
    tier = substr(label,lapply(label, function(x) max(unlist(gregexpr('!!',x)))+2),nchar(label))
  ) %>% 
  filter(tier %in% c('Speak English "not well"', 
                     'Speak English "not at all"', 
                     'Total', 'Speak Spanish', 
                     'Speak Asian and Pacific Island languages')) %>% 
  group_by(blockgroup, tier) %>% 
  summarise(
    estimate1 = sum(estimate)
  ) %>% 
  spread(
    key = "tier",
    value = "estimate1"
  ) %>% 
  mutate(
    `% speaking english < well` = (`Speak English "not well"` + `Speak English "not at all"`) / Total * 100,
    `% speaking english > well` = (100 - `% speaking english < well`),
    `% speaking spanish` = (`Speak Spanish`/ Total) * 100,
    `% not speaking spanish` = (100 - `% speaking spanish`),
    `% speaking api` = (`Speak Asian and Pacific Island languages` / Total) * 100
  ) %>% 
  left_join(sj_median_income_by_block %>% dplyr::select(-Median_Income)) %>% 
  filter(!is.na(device_count)) %>% 
  mutate(log_perc = log(`% speaking english < well`))

# loading in age data - specifically looking at percentage 65+ and percentage <30
sj_age_by_block <- getCensus(
    name = "acs/acs5",
    vintage = 2018,
    region = "block group:*", 
    regionin = "state:06+county:085",
    vars = "group(B01001)"
  ) %>% 
  mutate(
    blockgroup =
      paste0(state,county,tract,block_group)
  ) %>% 
  select_if(!names(.) %in% c("GEO_ID","state","county","tract","block_group","NAME")) %>% 
  select(-c(contains("EA"),contains("MA"),contains("M"))) %>% 
  gather(
    key = "variable",
    value = "estimate", 
    - blockgroup
  ) %>% 
  mutate(
    label = acs_vars$label[match(variable,acs_vars$name)]
  ) %>% 
  select(-variable) %>% 
  separate(
    label,
    into = c(NA,NA,"sex","age"),
    sep = "!!"
  ) %>% filter(!is.na(age)) %>% 
  mutate(elderly = ifelse(age %in% c("65 and 66 years", "67 to 69 years", "70 to 74 years", "75 to 79 years", "80 to 84 years", "85 years and over"), estimate, NA), `less than 30` = ifelse(age %in% c("Under 5 years", "5 to 9 years", "10 to 14 years", "15 to 17 years", "18 and 19 years", "20 years", "21 years", "22 to 24 years", "25 to 29 years"), estimate, NA)) %>% 
  group_by(blockgroup) %>% 
  summarize(elderly = sum(elderly, na.rm = T), `less than 30` = sum(`less than 30`, na.rm = T), total = sum(estimate, na.rm = T)) %>% 
  mutate(`percent elderly` = elderly*100 / total, `percent less than 30` = `less than 30`*100 / total, `percent nonelderly` = (100 - `percent elderly`)) %>% 
  left_join(sj_median_income_by_block %>% dplyr::select(-Median_Income)) %>% 
  filter(!is.na(device_count)) 

# get data on vehicles available as vehicles allocation
sj_vehicles_by_block <- getCensus(
    name = "acs/acs5",
    vintage = 2018,
    region = "block group:*", 
    regionin = "state:06+county:085",
    vars = "group(B992512)"
  ) %>% 
  mutate(
    blockgroup =
      paste0(state,county,tract,block_group)
  ) %>% 
  select_if(!names(.) %in% c("GEO_ID","state","county","tract","block_group","NAME")) %>% 
  select(-c(contains("EA"),contains("MA"),contains("M"))) %>% 
  dplyr::select(B992512_001E, blockgroup) %>%
  rename(total_vehicles = B992512_001E, blockgroup = blockgroup) %>%
  left_join(sj_age_by_block %>% dplyr::select_if(!names(.) %in% c("elderly", "percent elderly", "less than 30", "percent less than 30"))) %>%
  mutate(`vehicles per capita` = total_vehicles / total) %>%
  filter(!is.na(device_count)) 

# also get data on vehicles available as households without a vehicle
sj_no_vehicles_by_block <- getCensus(
    name = "acs/acs5",
    vintage = 2018,
    region = "block group:*", 
    regionin = "state:06+county:085",
    vars = "group(B25044)"
  ) %>% 
  mutate(
    blockgroup =
      paste0(state,county,tract,block_group)
  ) %>% 
  select_if(!names(.) %in% c("GEO_ID","state","county","tract","block_group","NAME")) %>% 
  select(-c(contains("EA"),contains("MA"),contains("M"))) %>% 
  gather(key = "variable", value = "estimate", -blockgroup) %>% 
  mutate(label = acs_vars$label[match(variable,acs_vars$name)]) %>% 
  select(-variable) %>%
  separate(label, into = c(NA, NA, NA,"vehicles"), sep = "!!") %>% 
  filter(!is.na(vehicles)) %>%
  group_by(blockgroup, vehicles) %>%
  summarize(grouped_vehicles = sum(estimate)) %>%
  spread(key = vehicles, value = grouped_vehicles) %>%
  mutate(total_nums = `1 vehicle available` + `2 vehicles available` + `3 vehicles available` + `4 vehicles available` + `5 or more vehicles available` + `No vehicle available`, `percent no vehicles` = `No vehicle available`*100 / total_nums, `percent with vehicles` = (100-`percent no vehicles`)) %>%
  left_join(sj_age_by_block %>% dplyr::select_if(!names(.) %in% c("elderly", "percent elderly", "less than 30", "percent less than 30"))) %>%
  filter(!is.na(device_count))

# get data on occupants per room
sj_occupants_per_room_by_block <- getCensus(
    name = "acs/acs5",
    vintage = 2018,
    region = "block group:*", 
    regionin = "state:06+county:085",
    vars = "group(B25014)"
  ) %>% 
  mutate(
    blockgroup =
      paste0(state,county,tract,block_group)
  ) %>% 
  select_if(!names(.) %in% c("GEO_ID","state","county","tract","block_group","NAME")) %>% 
  select(-c(contains("EA"),contains("MA"),contains("M"))) %>% 
  gather(key = "variable", value = "estimate", -blockgroup) %>% 
  mutate(label = acs_vars$label[match(variable,acs_vars$name)]) %>% 
  select(-variable) %>% 
  separate(label, into = c(NA, NA, NA,"occupants per room"), sep = "!!") %>% 
  filter(!is.na(`occupants per room`)) %>%
  group_by(blockgroup, `occupants per room`) %>%
  summarize(estimate_tot = sum(estimate)) %>% 
  spread(key = `occupants per room`, value = estimate_tot) %>%
  mutate(total_nums = `0.50 or less occupants per room` + `0.51 to 1.00 occupants per room` + `1.01 to 1.50 occupants per room` + `1.51 to 2.00 occupants per room` + `2.01 or more occupants per room`, `percent 1 or more` = (`1.01 to 1.50 occupants per room` + `1.51 to 2.00 occupants per room` + `2.01 or more occupants per room`) * 100/ total_nums, `percent less than 1` = (100-`percent 1 or more`)) %>%
  left_join(sj_age_by_block %>% dplyr::select_if(!names(.) %in% c("elderly", "percent elderly", "less than 30", "percent less than 30"))) %>%
  filter(!is.na(device_count)) 

Testing correlations

In the plots below, I show the selected variables against percent of devices completely at home since the shelter-in-place order started, as well as against percent of devices pre-shelter-in-place for comparison.

Age

# age
sj_age_by_block %>%
  ggplot(aes(
  x = `percent less than 30`,
  y = `% not completely at home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
labs(
    x = "Percent of residents younger than 30",
    y = "Percent devices leaving home on weekdays since shelter-in-place",
    title = "San Jose: Social Distancing and Young Age Groups"
  )

young_model <- lm(sj_age_by_block$`% not completely at home` ~ sj_age_by_block$`percent less than 30`)
summary(young_model)

## 
## Call:
## lm(formula = sj_age_by_block$`% not completely at home` ~ sj_age_by_block$`percent less than 30`)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -28.505  -4.978  -0.362   4.274  39.808 
## 
## Coefficients:
##                                        Estimate Std. Error t value Pr(>|t|)    
## (Intercept)                            44.49223    1.51348  29.397  < 2e-16 ***
## sj_age_by_block$`percent less than 30`  0.17973    0.03836   4.685 3.51e-06 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 8.299 on 567 degrees of freedom
## Multiple R-squared:  0.03726,    Adjusted R-squared:  0.03557 
## F-statistic: 21.95 on 1 and 567 DF,  p-value: 3.513e-06

sj_age_by_block %>% filter(`percent elderly` < 50) %>% # get rid of extreme outliers
  ggplot(aes(
  x = `percent elderly`,
  y = `% not completely at home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of residents 65 and older",
    y = "Percent devices leaving home on weekdays since shelter-in-place",
    title = "San Jose: Social Distancing and Elderly Population"
  )

elderly_model <- lm(`% not completely at home` ~ `percent elderly`, sj_age_by_block %>% filter(`percent elderly` < 50))
summary(elderly_model)

## 
## Call:
## lm(formula = `% not completely at home` ~ `percent elderly`, 
##     data = sj_age_by_block %>% filter(`percent elderly` < 50))
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -28.542  -5.120  -0.472   4.248  36.658 
## 
## Coefficients:
##                   Estimate Std. Error t value Pr(>|t|)    
## (Intercept)       53.65024    0.78297   68.52   <2e-16 ***
## `percent elderly` -0.17733    0.05406   -3.28   0.0011 ** 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 8.361 on 564 degrees of freedom
## Multiple R-squared:  0.01872,    Adjusted R-squared:  0.01698 
## F-statistic: 10.76 on 1 and 564 DF,  p-value: 0.001102

# compare this to pre-shelter-in-place behavior
sj_age_by_block %>%
  ggplot(aes(
  x = `percent less than 30`,
  y = `% not completely at home pre shelter`
)) + geom_point() + 
  geom_smooth(method=lm) + 
labs(
    x = "Percent of residents younger than 30",
    y = "Percent devices leaving home pre-shelter-in-place",
    title = "San Jose: Staying at Home and Young Age Groups Pre Shelter-in-Place"
  )

young_model2 <- lm(sj_age_by_block$`% not completely at home pre shelter` ~ sj_age_by_block$`percent less than 30`)
summary(young_model2)

## 
## Call:
## lm(formula = sj_age_by_block$`% not completely at home pre shelter` ~ 
##     sj_age_by_block$`percent less than 30`)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -28.1939  -2.8160  -0.1557   2.9950  16.7071 
## 
## Coefficients:
##                                        Estimate Std. Error t value Pr(>|t|)    
## (Intercept)                            81.87032    0.82253   99.53  < 2e-16 ***
## sj_age_by_block$`percent less than 30` -0.11072    0.02085   -5.31 1.57e-07 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 4.51 on 567 degrees of freedom
## Multiple R-squared:  0.04738,    Adjusted R-squared:  0.0457 
## F-statistic:  28.2 on 1 and 567 DF,  p-value: 1.573e-07

sj_age_by_block %>% filter(`percent elderly` < 50) %>% # get rid of extreme outliers
  ggplot(aes(
  x = `percent elderly`,
  y = `% not completely at home pre shelter`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of residents 65 and older",
    y = "Percent devices leaving home on weekdays pre-shelter-in-place",
    title = "San Jose: Staying at Home and Elderly Population Pre Shelter-in-Place"
  )

elderly_model2 <- lm(`% not completely at home pre shelter` ~ `percent elderly`, sj_age_by_block %>% filter(`percent elderly` < 50))
summary(elderly_model2)

## 
## Call:
## lm(formula = `% not completely at home pre shelter` ~ `percent elderly`, 
##     data = sj_age_by_block %>% filter(`percent elderly` < 50))
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -27.236  -2.830  -0.158   3.145  14.296 
## 
## Coefficients:
##                   Estimate Std. Error t value Pr(>|t|)    
## (Intercept)        75.9045     0.4257 178.295  < 2e-16 ***
## `percent elderly`   0.1329     0.0294   4.522 7.47e-06 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 4.546 on 564 degrees of freedom
## Multiple R-squared:  0.03499,    Adjusted R-squared:  0.03328 
## F-statistic: 20.45 on 1 and 564 DF,  p-value: 7.466e-06

Income

# income - less than $75000
sj_ami_by_block %>% 
  ggplot(aes(
  x = `% over 75,000`,
  y = `% not completely at home`
)) + geom_point() + 
  geom_smooth(method=lm) +
  labs(
    x = "Percent of housholds with incomes over $75,000 (50% AMI) annually",
    y = "Percent devices leaving home on weekdays since shelter-in-place",
    title = "San Jose: Social Distancing and Households Above 50% AMI"
  )

income_75_model <- lm(`% not completely at home` ~ `% over 75,000`, sj_ami_by_block)
summary(income_75_model)

## 
## Call:
## lm(formula = `% not completely at home` ~ `% over 75,000`, data = sj_ami_by_block)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -21.389  -4.795  -0.606   4.185  34.925 
## 
## Coefficients:
##                 Estimate Std. Error t value Pr(>|t|)    
## (Intercept)     64.35928    1.11518   57.71   <2e-16 ***
## `% over 75,000` -0.20909    0.01719  -12.16   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 7.444 on 566 degrees of freedom
##   (1 observation deleted due to missingness)
## Multiple R-squared:  0.2072, Adjusted R-squared:  0.2058 
## F-statistic:   148 on 1 and 566 DF,  p-value: < 2.2e-16

# income - less than $100000
sj_ami_by_block %>% 
  ggplot(aes(
  x = `% over 100,000`,
  y = `% not completely at home`
)) + geom_point() + 
  geom_smooth(method=lm) +
  labs(
    x = "Percent of housholds with incomes over $100,000 (80% AMI) annually",
    y = "Percent devices leaving home on weekdays since shelter-in-place",
    title = "San Jose: Social Distancing and Households Below 80% AMI"
  )

income_100_model <- lm(`% not completely at home` ~ `% over 100,000`, sj_ami_by_block)
summary(income_100_model)

## 
## Call:
## lm(formula = `% not completely at home` ~ `% over 100,000`, data = sj_ami_by_block)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -20.762  -4.639  -0.500   4.176  33.309 
## 
## Coefficients:
##                  Estimate Std. Error t value Pr(>|t|)    
## (Intercept)      61.78423    0.87244   70.82   <2e-16 ***
## `% over 100,000` -0.20475    0.01599  -12.80   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 7.362 on 566 degrees of freedom
##   (1 observation deleted due to missingness)
## Multiple R-squared:  0.2246, Adjusted R-squared:  0.2232 
## F-statistic: 163.9 on 1 and 566 DF,  p-value: < 2.2e-16

# income - less than $125000
sj_ami_by_block %>% 
  ggplot(aes(
  x = `% over 125,000`,
  y = `% not completely at home`
)) + geom_point() + 
  geom_smooth(method=lm) +
  labs(
    x = "Percent of housholds with incomes over $125,000 annually",
    y = "Percent devices leaving home on weekdays since shelter-in-place",
    title = "San Jose: Social Distancing and Households Below $125,000"
  )

income_125_model <- lm(`% not completely at home` ~ `% over 125,000`, sj_ami_by_block)
summary(income_125_model)

## 
## Call:
## lm(formula = `% not completely at home` ~ `% over 125,000`, data = sj_ami_by_block)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -23.072  -4.669  -0.845   4.018  32.374 
## 
## Coefficients:
##                  Estimate Std. Error t value Pr(>|t|)    
## (Intercept)      60.00338    0.73524   81.61   <2e-16 ***
## `% over 125,000` -0.21065    0.01623  -12.98   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 7.339 on 566 degrees of freedom
##   (1 observation deleted due to missingness)
## Multiple R-squared:  0.2294, Adjusted R-squared:  0.228 
## F-statistic: 168.5 on 1 and 566 DF,  p-value: < 2.2e-16

# compare to pre shelter in place
sj_ami_by_block %>% 
  ggplot(aes(
  x = `% over 75,000`,
  y = `% not completely at home pre shelter`
)) + geom_point() + 
  geom_smooth(method=lm) +
  labs(
    x = "Percent of housholds with incomes over $75,000 (50% AMI) annually",
    y = "Percent devices leaving home on weekdays pre-shelter-in-place",
    title = "San Jose: Staying at Home and Households Above 50% AMI Pre Shelter-in-Place"
  )

income_75_model2 <- lm(`% not completely at home pre shelter` ~ `% over 75,000`, sj_ami_by_block)
summary(income_75_model2)

## 
## Call:
## lm(formula = `% not completely at home pre shelter` ~ `% over 75,000`, 
##     data = sj_ami_by_block)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -28.4357  -2.7003  -0.1437   2.7764  16.6680 
## 
## Coefficients:
##                 Estimate Std. Error t value Pr(>|t|)    
## (Intercept)     72.61447    0.65712 110.504  < 2e-16 ***
## `% over 75,000`  0.08029    0.01013   7.926 1.21e-14 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 4.386 on 566 degrees of freedom
##   (1 observation deleted due to missingness)
## Multiple R-squared:  0.09991,    Adjusted R-squared:  0.09832 
## F-statistic: 62.83 on 1 and 566 DF,  p-value: 1.206e-14

# income - less than $100000
sj_ami_by_block %>% 
  ggplot(aes(
  x = `% over 100,000`,
  y = `% not completely at home pre shelter`
)) + geom_point() + 
  geom_smooth(method=lm) +
  labs(
    x = "Percent of housholds with incomes over $100,000 (80% AMI) annually",
    y = "Percent devices leaving home on weekdays pre-shelter-in-place",
    title = "San Jose: Staying Home and Households Below 80% AMI Pre Shelter-in-Place"
  )

income_100_model2 <- lm(`% not completely at home pre shelter` ~ `% over 100,000`, sj_ami_by_block)
summary(income_100_model2)

## 
## Call:
## lm(formula = `% not completely at home pre shelter` ~ `% over 100,000`, 
##     data = sj_ami_by_block)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -28.5034  -2.6406   0.0803   2.5599  16.9387 
## 
## Coefficients:
##                  Estimate Std. Error t value Pr(>|t|)    
## (Intercept)      73.26132    0.51177 143.152   <2e-16 ***
## `% over 100,000`  0.08532    0.00938   9.096   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 4.319 on 566 degrees of freedom
##   (1 observation deleted due to missingness)
## Multiple R-squared:  0.1275, Adjusted R-squared:  0.126 
## F-statistic: 82.73 on 1 and 566 DF,  p-value: < 2.2e-16

# over 125000
sj_ami_by_block %>% 
  ggplot(aes(
  x = `% over 125,000`,
  y = `% not completely at home pre shelter`
)) + geom_point() + 
  geom_smooth(method=lm) +
  labs(
    x = "Percent of housholds with incomes over $125,000 annually",
    y = "Percent devices leaving home on weekdays pre-shelter-in-place",
    title = "San Jose: Social Distancing and Households Below $125,000 Pre Shelter-in-Place"
  )

income_125_model2 <- lm(`% not completely at home pre shelter` ~ `% over 125,000`, sj_ami_by_block)
summary(income_125_model2)

## 
## Call:
## lm(formula = `% not completely at home pre shelter` ~ `% over 125,000`, 
##     data = sj_ami_by_block)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -28.353  -2.556   0.022   2.522  16.560 
## 
## Coefficients:
##                   Estimate Std. Error t value Pr(>|t|)    
## (Intercept)      73.640242   0.425069   173.2   <2e-16 ***
## `% over 125,000`  0.096607   0.009382    10.3   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 4.243 on 566 degrees of freedom
##   (1 observation deleted due to missingness)
## Multiple R-squared:  0.1578, Adjusted R-squared:  0.1563 
## F-statistic:   106 on 1 and 566 DF,  p-value: < 2.2e-16

Language

# language
sj_lang_by_block %>% 
  ggplot(aes(
  x = `% speaking english > well`,
  y = `% not completely at home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of individuals speaking English well",
    y = "Percent devices leaving home on weekdays since shelter-in-place",
    title = "San Jose: Social Distancing and English Language Ability"
  )

english_ability_model <- lm(`% not completely at home` ~ `% speaking english > well`, sj_lang_by_block)
summary(english_ability_model)

## 
## Call:
## lm(formula = `% not completely at home` ~ `% speaking english > well`, 
##     data = sj_lang_by_block)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -25.467  -5.070  -0.542   3.887  39.230 
## 
## Coefficients:
##                             Estimate Std. Error t value Pr(>|t|)    
## (Intercept)                 67.28491    3.36107  20.019  < 2e-16 ***
## `% speaking english > well` -0.17915    0.03769  -4.754 2.53e-06 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 8.294 on 567 degrees of freedom
## Multiple R-squared:  0.03833,    Adjusted R-squared:  0.03663 
## F-statistic:  22.6 on 1 and 567 DF,  p-value: 2.534e-06

sj_lang_by_block %>% 
  ggplot(aes(
  x = `% not speaking spanish`,
  y = `% not completely at home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of individuals not speaking Spanish",
    y = "Percent devices leaving home on weekdays since shelter-in-place",
    title = "San Jose: Social Distancing and Spanish Language Ability"
  )

spanish_speaking_model <- lm(`% not completely at home` ~ `% not speaking spanish`, sj_lang_by_block)
summary(spanish_speaking_model)

## 
## Call:
## lm(formula = `% not completely at home` ~ `% not speaking spanish`, 
##     data = sj_lang_by_block)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -25.434  -4.585  -0.796   3.568  38.341 
## 
## Coefficients:
##                          Estimate Std. Error t value Pr(>|t|)    
## (Intercept)              63.61990    1.32137  48.147   <2e-16 ***
## `% not speaking spanish` -0.15727    0.01646  -9.555   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 7.849 on 567 degrees of freedom
## Multiple R-squared:  0.1387, Adjusted R-squared:  0.1372 
## F-statistic: 91.29 on 1 and 567 DF,  p-value: < 2.2e-16

# compare to pre shelter in place
sj_lang_by_block %>% 
  ggplot(aes(
  x = `% speaking english > well`,
  y = `% not completely at home pre shelter`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of individuals speaking English well",
    y = "Percent devices leaving home on weekdays pre-shelter-in-place",
    title = "San Jose: Staying at Home and English Language Ability Pre Shelter-in-Place"
  )

english_ability_model2 <- lm(`% not completely at home pre shelter` ~ `% speaking english > well`, sj_lang_by_block)
summary(english_ability_model2)

## 
## Call:
## lm(formula = `% not completely at home pre shelter` ~ `% speaking english > well`, 
##     data = sj_lang_by_block)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -28.9364  -2.4342   0.0388   3.0316  12.3011 
## 
## Coefficients:
##                             Estimate Std. Error t value Pr(>|t|)    
## (Intercept)                 60.84131    1.73337  35.100   <2e-16 ***
## `% speaking english > well`  0.18913    0.01943   9.732   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 4.277 on 567 degrees of freedom
## Multiple R-squared:  0.1431, Adjusted R-squared:  0.1416 
## F-statistic:  94.7 on 1 and 567 DF,  p-value: < 2.2e-16

sj_lang_by_block %>% 
  ggplot(aes(
  x = `% not speaking spanish`,
  y = `% not completely at home pre shelter`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of individuals not speaking Spanish",
    y = "Percent devices leaving home on weekdays pre shelter-in-place",
    title = "San Jose: Staying at Home and Spanish Language Ability Pre Shelter-in-Place"
  )

spanish_speaking_model2 <- lm(`% not completely at home pre shelter` ~ `% not speaking spanish`, sj_lang_by_block)
summary(spanish_speaking_model2)

## 
## Call:
## lm(formula = `% not completely at home pre shelter` ~ `% not speaking spanish`, 
##     data = sj_lang_by_block)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -28.793  -2.540  -0.002   2.680  11.988 
## 
## Coefficients:
##                           Estimate Std. Error t value Pr(>|t|)    
## (Intercept)              71.228200   0.726831  97.998   <2e-16 ***
## `% not speaking spanish`  0.082206   0.009054   9.079   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 4.318 on 567 degrees of freedom
## Multiple R-squared:  0.1269, Adjusted R-squared:  0.1254 
## F-statistic: 82.43 on 1 and 567 DF,  p-value: < 2.2e-16

Occupants per room

# occupants per room
sj_occupants_per_room_by_block %>% 
  ggplot(aes(
  x = `percent less than 1`,
  y = `% not completely at home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of households with 1 or fewer occupant per room",
    y = "Percent devices leaving home on weekdays since shelter-in-place",
    title = "San Jose: Social Distancing and Room Occupancy"
  )

occupants_model <- lm(`% not completely at home` ~ `percent less than 1`, sj_occupants_per_room_by_block)
summary(occupants_model)

## 
## Call:
## lm(formula = `% not completely at home` ~ `percent less than 1`, 
##     data = sj_occupants_per_room_by_block)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -24.657  -4.937  -0.288   3.901  35.043 
## 
## Coefficients:
##                       Estimate Std. Error t value Pr(>|t|)    
## (Intercept)           70.49553    2.95264  23.875  < 2e-16 ***
## `percent less than 1` -0.21239    0.03252  -6.532 1.45e-10 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 8.062 on 566 degrees of freedom
##   (1 observation deleted due to missingness)
## Multiple R-squared:  0.07009,    Adjusted R-squared:  0.06845 
## F-statistic: 42.66 on 1 and 566 DF,  p-value: 1.451e-10

# compare to pre shelter in place
sj_occupants_per_room_by_block %>% 
  ggplot(aes(
  x = `percent less than 1`,
  y = `% not completely at home pre shelter`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of households with 1 or fewer occupant per room",
    y = "Percent devices leaving home on weekdays pre shelter-in-place",
    title = "San Jose: Staying at Home and Room Occupancy Pre Shelter-in-Place"
  )

occupants_model2 <- lm(`% not completely at home pre shelter` ~ `percent less than 1`, sj_occupants_per_room_by_block)
summary(occupants_model2)

## 
## Call:
## lm(formula = `% not completely at home pre shelter` ~ `percent less than 1`, 
##     data = sj_occupants_per_room_by_block)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -28.3246  -2.6506  -0.2808   2.7536  17.0509 
## 
## Coefficients:
##                       Estimate Std. Error t value Pr(>|t|)    
## (Intercept)           62.88485    1.57437  39.943   <2e-16 ***
## `percent less than 1`  0.16329    0.01734   9.418   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 4.299 on 566 degrees of freedom
##   (1 observation deleted due to missingness)
## Multiple R-squared:  0.1355, Adjusted R-squared:  0.134 
## F-statistic:  88.7 on 1 and 566 DF,  p-value: < 2.2e-16

Vehicle ownership

# vehicles
sj_vehicles_by_block %>% 
  ggplot(aes(
  x = `vehicles per capita`,
  y = `% not completely at home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Number of vehicles per capita",
    y = "Percent devices leaving home on weekdays since shelter-in-place",
    title = "San Jose: Social Distancing and Vehicles Per Capita"
  )

# vehicles - percent with no vehicles
sj_no_vehicles_by_block %>% 
  ggplot(aes(
  x = `percent with vehicles`,
  y = `% not completely at home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of housholds with vehicles available",
    y = "Percent devices leaving home on weekdays since shelter-in-place",
    title = "San Jose: Social Distancing and Vehicle Availability"
  )

vehicles_model <- lm(`% not completely at home` ~ `percent with vehicles`, sj_no_vehicles_by_block)
summary(vehicles_model)

## 
## Call:
## lm(formula = `% not completely at home` ~ `percent with vehicles`, 
##     data = sj_no_vehicles_by_block)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -25.500  -5.226  -0.406   4.579  38.500 
## 
## Coefficients:
##                         Estimate Std. Error t value Pr(>|t|)    
## (Intercept)             75.71497    5.14420   14.72  < 2e-16 ***
## `percent with vehicles` -0.25615    0.05393   -4.75 2.59e-06 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 8.199 on 566 degrees of freedom
##   (1 observation deleted due to missingness)
## Multiple R-squared:  0.03833,    Adjusted R-squared:  0.03663 
## F-statistic: 22.56 on 1 and 566 DF,  p-value: 2.587e-06

# compare to pre shelter in place
sj_no_vehicles_by_block %>% 
  ggplot(aes(
  x = `percent with vehicles`,
  y = `% not completely at home pre shelter`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of housholds with vehicles available",
    y = "Percent devices leaving home on weekdays pre shelter-in-place",
    title = "San Jose: Social Distancing and Vehicle Availability Pre Shelter-in-Place"
  )

vehicles_model2 <- lm(`% not completely at home pre shelter` ~ `percent with vehicles`, sj_no_vehicles_by_block)
summary(vehicles_model2)

## 
## Call:
## lm(formula = `% not completely at home pre shelter` ~ `percent with vehicles`, 
##     data = sj_no_vehicles_by_block)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -25.5618  -2.9606  -0.0694   3.0006  12.6053 
## 
## Coefficients:
##                         Estimate Std. Error t value Pr(>|t|)    
## (Intercept)             63.25942    2.83717  22.297  < 2e-16 ***
## `percent with vehicles`  0.15084    0.02975   5.071 5.37e-07 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 4.522 on 566 degrees of freedom
##   (1 observation deleted due to missingness)
## Multiple R-squared:  0.04346,    Adjusted R-squared:  0.04177 
## F-statistic: 25.72 on 1 and 566 DF,  p-value: 5.37e-07

Additions based on literature search

I also consider education and internet access, based on previous research. ## Education

sj_education_by_block <- getCensus(
    name = "acs/acs5",
    vintage = 2018,
    region = "block group:*", 
    regionin = "state:06+county:085",
    vars = "group(B15003)"
  ) %>% 
  mutate(
    blockgroup =
      paste0(state,county,tract,block_group)
  ) %>% 
  select_if(!names(.) %in% c("GEO_ID","state","county","tract","block_group","NAME")) %>% 
  select(-c(contains("EA"),contains("MA"),contains("M"))) %>% 
  gather(key = "variable", value = "estimate", -blockgroup) %>% 
  mutate(label = acs_vars$label[match(variable,acs_vars$name)]) %>% 
  select(-variable) %>% 
  separate(label, into = c(NA, NA, "education level"), sep = "!!") %>% 
  mutate(`education level` = replace_na(`education level`, "total_educ")) %>% # if the education level field is NA, this corresponded to the total number in that blockgroup
  spread(key = `education level`, value = estimate) %>%
  mutate(`percent associates or higher` = (`Associate's degree` + `Bachelor's degree` + `Doctorate degree` + `Master's degree`)*100/total_educ, `percent less than associates` = 100-`percent associates or higher`) %>% 
  left_join(sj_age_by_block %>% dplyr::select_if(!names(.) %in% c("elderly", "percent elderly", "less than 30", "percent less than 30"))) %>%
  filter(!is.na(device_count)) 

# plotting
sj_education_by_block %>% 
  ggplot(aes(
  x = `percent associates or higher`,
  y = `% not completely at home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of people with an degree at Associate's level or higher",
    y = "Percent devices leaving home on weekdays since shelter-in-place",
    title = "San Jose: Social Distancing and Education"
  )

educ_model <- lm(`% not completely at home` ~ `percent associates or higher`, sj_education_by_block)
summary(educ_model)

## 
## Call:
## lm(formula = `% not completely at home` ~ `percent associates or higher`, 
##     data = sj_education_by_block)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -23.619  -4.747  -0.950   3.300  43.002 
## 
## Coefficients:
##                                Estimate Std. Error t value Pr(>|t|)    
## (Intercept)                     60.4003     0.8408   71.84   <2e-16 ***
## `percent associates or higher`  -0.1910     0.0165  -11.58   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 7.606 on 567 degrees of freedom
## Multiple R-squared:  0.1912, Adjusted R-squared:  0.1898 
## F-statistic: 134.1 on 1 and 567 DF,  p-value: < 2.2e-16

# compare to pre shelter in place
sj_education_by_block %>% 
  ggplot(aes(
  x = `percent associates or higher`,
  y = `% not completely at home pre shelter`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of people with an degree at Associate's level or higher",
    y = "Percent devices leaving home on weekdays pre-shelter-in-place",
    title = "San Jose: Social Distancing and Education Pre Shelter-in-Place"
  )

educ_model2 <- lm(`% not completely at home pre shelter` ~ `percent associates or higher`, sj_education_by_block)
summary(educ_model2)

## 
## Call:
## lm(formula = `% not completely at home pre shelter` ~ `percent associates or higher`, 
##     data = sj_education_by_block)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -30.2224  -2.5458   0.0672   2.7859  13.9180 
## 
## Coefficients:
##                                 Estimate Std. Error t value Pr(>|t|)    
## (Intercept)                    73.547853   0.476189  154.45   <2e-16 ***
## `percent associates or higher`  0.086343   0.009344    9.24   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 4.308 on 567 degrees of freedom
## Multiple R-squared:  0.1309, Adjusted R-squared:  0.1293 
## F-statistic: 85.39 on 1 and 567 DF,  p-value: < 2.2e-16

Internet

sj_internet_by_block <- getCensus(
    name = "acs/acs5",
    vintage = 2018,
    region = "block group:*", 
    regionin = "state:06+county:085",
    vars = "group(B28002)"
  ) %>% 
  mutate(
    blockgroup =
      paste0(state,county,tract,block_group)
  ) %>% 
  select_if(!names(.) %in% c("GEO_ID","state","county","tract","block_group","NAME")) %>% 
  select(-c(contains("EA"),contains("MA"),contains("M"))) %>% 
  gather(key = "variable", value = "estimate", -blockgroup) %>% 
  mutate(label = acs_vars$label[match(variable,acs_vars$name)]) %>% 
  select(-variable) %>% 
  separate(label, into = c(NA, NA, "subscription", "type", "additional"), sep = "!!") %>% 
  filter(is.na(subscription) | (type == "Broadband such as cable, fiber optic or DSL") & is.na(additional)) %>% 
  mutate(type = replace_na(type, "total_num")) %>% 
  dplyr::select(blockgroup, type, estimate) %>%
  spread(key = type, value = estimate) %>%
  mutate(`percent high speed` = `Broadband such as cable, fiber optic or DSL`*100/total_num, `percent no high speed` = 100-`percent high speed`) %>% 
  left_join(sj_age_by_block %>% dplyr::select_if(!names(.) %in% c("elderly", "percent elderly", "less than 30", "percent less than 30"))) %>%
  filter(!is.na(device_count)) 

# plotting
sj_internet_by_block %>% 
  ggplot(aes(
  x = `percent high speed`,
  y = `% not completely at home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of households with broadband such as cable, fiber optic or DSL",
    y = "Percent devices leaving home on weekdays since shelter-in-place",
    title = "San Jose: Social Distancing and High Speed Internet"
  )

internet_model <- lm(`% not completely at home` ~ `percent high speed`, sj_internet_by_block)
summary(internet_model)

## 
## Call:
## lm(formula = `% not completely at home` ~ `percent high speed`, 
##     data = sj_internet_by_block)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -21.884  -4.676  -0.549   3.996  38.086 
## 
## Coefficients:
##                      Estimate Std. Error t value Pr(>|t|)    
## (Intercept)          72.72612    2.22683  32.659   <2e-16 ***
## `percent high speed` -0.26512    0.02731  -9.709   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 7.741 on 566 degrees of freedom
##   (1 observation deleted due to missingness)
## Multiple R-squared:  0.1428, Adjusted R-squared:  0.1413 
## F-statistic: 94.26 on 1 and 566 DF,  p-value: < 2.2e-16

# compare to pre-shelter-in-place behavior
sj_internet_by_block %>% 
  ggplot(aes(
  x = `percent high speed`,
  y = `% not completely at home pre shelter`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of households with broadband such as cable, fiber optic or DSL",
    y = "Percent devices leaving home on weekdays pre-shelter-in-place",
    title = "San Jose: Social Distancing and High Speed Internet Pre Shelter-in-Place"
  )

internet_model2 <- lm(`% not completely at home pre shelter` ~ `percent high speed`, sj_internet_by_block)
summary(internet_model2)

## 
## Call:
## lm(formula = `% not completely at home pre shelter` ~ `percent high speed`, 
##     data = sj_internet_by_block)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -28.5915  -2.8440  -0.1157   2.8288  16.8750 
## 
## Coefficients:
##                      Estimate Std. Error t value Pr(>|t|)    
## (Intercept)           69.1067     1.2800  53.990  < 2e-16 ***
## `percent high speed`   0.1055     0.0157   6.719 4.48e-11 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 4.449 on 566 degrees of freedom
##   (1 observation deleted due to missingness)
## Multiple R-squared:  0.07386,    Adjusted R-squared:  0.07223 
## F-statistic: 45.14 on 1 and 566 DF,  p-value: 4.477e-11

Sex of workers

# use LODES data
sj_sex_workers_by_block <- grab_lodes(
    state = "ca", 
    year = 2017, 
    lodes_type = "rac", 
    job_type = "JT01",
    segment = "S000", 
    state_part = "main",
    agg_geo = "bg"
  ) %>%
  left_join(sj_blockgroups, by = c("h_bg" = "GEOID")) %>% 
  filter(!is.na(DISTRICTS)) %>%
  dplyr::select(CS01, CS02, h_bg) %>%
  rename(male = CS01, female = CS02, blockgroup = h_bg) %>%
  mutate(total_workers = male + female, `% male workers` = male*100 / total_workers) %>%
  left_join(sj_age_by_block %>% dplyr::select_if(!names(.) %in% c("elderly", "percent elderly", "less than 30", "percent less than 30"))) %>%
  filter(!is.na(device_count))
  
# plotting
sj_sex_workers_by_block %>% 
  ggplot(aes(
  x = `% male workers`,
  y = `% not completely at home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of workers that are male",
    y = "Percent devices leaving home on weekdays since shelter-in-place",
    title = "San Jose: Social Distancing and Sex of Workers"
  )

sex_workers_model <- lm(`% not completely at home` ~ `% male workers`, sj_sex_workers_by_block)
summary(sex_workers_model)

## 
## Call:
## lm(formula = `% not completely at home` ~ `% male workers`, data = sj_sex_workers_by_block)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -25.516  -5.392  -0.510   4.469  36.365 
## 
## Coefficients:
##                  Estimate Std. Error t value Pr(>|t|)    
## (Intercept)       67.2451     7.1783   9.368   <2e-16 ***
## `% male workers`  -0.2951     0.1338  -2.206   0.0278 *  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 8.388 on 566 degrees of freedom
## Multiple R-squared:  0.008526,   Adjusted R-squared:  0.006774 
## F-statistic: 4.867 on 1 and 566 DF,  p-value: 0.02777

# compare to pre-shelter-in-place behavior
sj_sex_workers_by_block %>% 
  ggplot(aes(
  x = `% male workers`,
  y = `% not completely at home pre shelter`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of workers that are male",
    y = "Percent devices leaving home on weekdays pre-shelter-in-place",
    title = "San Jose: Social Distancing and Sex of Workers Pre Shelter-in-Place"
  )

sex_workers_model_2 <- lm(`% not completely at home pre shelter` ~ `% male workers`, sj_sex_workers_by_block)
summary(sex_workers_model_2)

## 
## Call:
## lm(formula = `% not completely at home pre shelter` ~ `% male workers`, 
##     data = sj_sex_workers_by_block)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -27.5200  -3.1184  -0.1399   3.0954  12.4701 
## 
## Coefficients:
##                  Estimate Std. Error t value Pr(>|t|)    
## (Intercept)      78.61111    3.95446  19.879   <2e-16 ***
## `% male workers` -0.01836    0.07369  -0.249    0.803    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 4.621 on 566 degrees of freedom
## Multiple R-squared:  0.0001096,  Adjusted R-squared:  -0.001657 
## F-statistic: 0.06207 on 1 and 566 DF,  p-value: 0.8033

This does not seem to be significant.

Race

sj_race_by_block <- getCensus(
    name = "acs/acs5",
    vintage = 2018,
    region = "block group:*", 
    regionin = "state:06+county:085",
    vars = "group(B02001)"
  ) %>% 
  mutate(
    blockgroup =
      paste0(state,county,tract,block_group)
  ) %>% 
  select_if(!names(.) %in% c("GEO_ID","state","county","tract","block_group","NAME")) %>% 
  select(-c(contains("EA"),contains("MA"),contains("M"))) %>% 
  gather(key = "variable", value = "estimate", -blockgroup) %>% 
  mutate(label = acs_vars$label[match(variable,acs_vars$name)]) %>% 
  select(-variable) %>% 
  separate(label, into = c(NA, NA, "race", "specification"), sep = "!!") %>% 
  filter(is.na(specification) & !is.na(race)) %>% 
  dplyr::select(blockgroup, estimate, race) %>%
  spread(key = race, value = estimate) %>% 
  mutate(total_race = `American Indian and Alaska Native alone` + `Asian alone` + `Black or African American alone` + `Native Hawaiian and Other Pacific Islander alone` + `Some other race alone` + `Two or more races` + `White alone`, `% white` = `White alone`*100/total_race, `% Asian` = `Asian alone`*100/total_race, `% black` = `Black or African American alone`*100/total_race) %>%  
  left_join(sj_age_by_block %>% dplyr::select_if(!names(.) %in% c("elderly", "percent elderly", "less than 30", "percent less than 30"))) %>%
  filter(!is.na(device_count)) 

# also get ethnicity data (hispanic/latino vs not)
sj_hisplat_by_block <- getCensus(
    name = "acs/acs5",
    vintage = 2018,
    region = "block group:*", 
    regionin = "state:06+county:085",
    vars = "group(B03002)"
  ) %>% 
  mutate(
    blockgroup =
      paste0(state,county,tract,block_group)
  ) %>% 
  select_if(!names(.) %in% c("GEO_ID","state","county","tract","block_group","NAME")) %>% 
  select(-c(contains("EA"),contains("MA"),contains("M"))) %>% 
  gather(key = "variable", value = "estimate", -blockgroup) %>% 
  mutate(label = acs_vars$label[match(variable,acs_vars$name)]) %>% 
  select(-variable) %>% 
  separate(label, into = c(NA, NA, "hisp/lat", "specification"), sep = "!!") %>%
  filter(is.na(specification) & !is.na(`hisp/lat`)) %>% 
  dplyr::select(blockgroup, estimate, `hisp/lat`) %>% 
  spread(key = `hisp/lat`, value = estimate) %>%
  mutate(`% non hispanic/latino` = `Not Hispanic or Latino`*100/(`Hispanic or Latino` + `Not Hispanic or Latino`))

# join with the race data
sj_race_by_block <- sj_race_by_block %>% left_join(sj_hisplat_by_block %>% dplyr::select(blockgroup, `% non hispanic/latino`))

# plotting
# percent white
sj_race_by_block %>% 
  ggplot(aes(
  x = `% white`,
  y = `% not completely at home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of residents that are white",
    y = "Percent devices leaving home on weekdays since shelter-in-place",
    title = "San Jose: Social Distancing and White Residents"
  )

white_model <- lm(`% not completely at home` ~ `% white`, sj_race_by_block)
summary(white_model)

## 
## Call:
## lm(formula = `% not completely at home` ~ `% white`, data = sj_race_by_block)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -25.625  -5.270  -0.390   4.665  37.436 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)    
## (Intercept) 50.12965    0.78574    63.8   <2e-16 ***
## `% white`    0.02933    0.01630     1.8   0.0724 .  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 8.434 on 567 degrees of freedom
## Multiple R-squared:  0.005681,   Adjusted R-squared:  0.003927 
## F-statistic:  3.24 on 1 and 567 DF,  p-value: 0.07241

# compare to pre-shelter-in-place behavior
sj_race_by_block %>% 
  ggplot(aes(
  x = `% white`,
  y = `% not completely at home pre shelter`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of residents that are white",
    y = "Percent devices leaving home on weekdays pre-shelter-in-place",
    title = "San Jose: Social Distancing and White Residents Pre Shelter-in-Place"
  )

white_model2 <- lm(`% not completely at home pre shelter` ~ `% white`, sj_race_by_block)
summary(white_model2)

## 
## Call:
## lm(formula = `% not completely at home pre shelter` ~ `% white`, 
##     data = sj_race_by_block)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -28.2995  -2.8302  -0.0347   2.8574  12.5621 
## 
## Coefficients:
##              Estimate Std. Error t value Pr(>|t|)    
## (Intercept) 74.834803   0.410114 182.473  < 2e-16 ***
## `% white`    0.064673   0.008506   7.603 1.21e-13 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 4.402 on 567 degrees of freedom
## Multiple R-squared:  0.09251,    Adjusted R-squared:  0.09091 
## F-statistic:  57.8 on 1 and 567 DF,  p-value: 1.209e-13

# percent Asian
sj_race_by_block %>% 
  ggplot(aes(
  x = `% Asian`,
  y = `% not completely at home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of residents that are Asian",
    y = "Percent devices leaving home on weekdays since shelter-in-place",
    title = "San Jose: Social Distancing and Asian Residents"
  )

asian_model <- lm(`% not completely at home` ~ `% Asian`, sj_race_by_block)
summary(asian_model)

## 
## Call:
## lm(formula = `% not completely at home` ~ `% Asian`, data = sj_race_by_block)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -21.032  -5.012  -0.564   4.386  36.002 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)    
## (Intercept) 56.20411    0.59228  94.894   <2e-16 ***
## `% Asian`   -0.14825    0.01519  -9.757   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 7.826 on 567 degrees of freedom
## Multiple R-squared:  0.1438, Adjusted R-squared:  0.1423 
## F-statistic: 95.21 on 1 and 567 DF,  p-value: < 2.2e-16

# compare to pre-shelter-in-place behavior
sj_race_by_block %>% 
  ggplot(aes(
  x = `% Asian`,
  y = `% not completely at home pre shelter`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of residents that are Asian",
    y = "Percent devices leaving home on weekdays pre-shelter-in-place",
    title = "San Jose: Social Distancing and Asian Residents Pre Shelter-in-Place"
  )

asian_model2 <- lm(`% not completely at home pre shelter` ~ `% Asian`, sj_race_by_block)
summary(asian_model2)

## 
## Call:
## lm(formula = `% not completely at home pre shelter` ~ `% Asian`, 
##     data = sj_race_by_block)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -27.6447  -3.0068   0.0283   3.2005  12.1305 
## 
## Coefficients:
##              Estimate Std. Error t value Pr(>|t|)    
## (Intercept) 78.069520   0.348967  223.72   <2e-16 ***
## `% Asian`   -0.013871   0.008952   -1.55    0.122    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 4.611 on 567 degrees of freedom
## Multiple R-squared:  0.004217,   Adjusted R-squared:  0.00246 
## F-statistic: 2.401 on 1 and 567 DF,  p-value: 0.1218

# percent non hispanic/latino
sj_race_by_block %>% 
  ggplot(aes(
  x = `% non hispanic/latino`,
  y = `% not completely at home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of residents that are not Hispanic or Latino",
    y = "Percent devices leaving home on weekdays since shelter-in-place",
    title = "San Jose: Social Distancing and Hispanic/Latino Residents"
  )

hisp_model <- lm(`% not completely at home` ~ `% non hispanic/latino`, sj_race_by_block)
summary(hisp_model)

## 
## Call:
## lm(formula = `% not completely at home` ~ `% non hispanic/latino`, 
##     data = sj_race_by_block)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -23.581  -4.752  -0.802   3.749  37.993 
## 
## Coefficients:
##                         Estimate Std. Error t value Pr(>|t|)    
## (Intercept)             62.21783    1.02147   60.91   <2e-16 ***
## `% non hispanic/latino` -0.15911    0.01425  -11.16   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 7.658 on 567 degrees of freedom
## Multiple R-squared:  0.1802, Adjusted R-squared:  0.1787 
## F-statistic: 124.6 on 1 and 567 DF,  p-value: < 2.2e-16

# compare to pre-shelter-in-place behavior
sj_race_by_block %>% 
  ggplot(aes(
  x = `% non hispanic/latino`,
  y = `% not completely at home pre shelter`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of residents that are not Hispanic or Latino",
    y = "Percent devices leaving home on weekdays pre-shelter-in-place",
    title = "San Jose: Social Distancing and Hispanic/Latino Residents Pre Shelter-in-Place"
  )

hisp_model2 <- lm(`% not completely at home pre shelter` ~ `% non hispanic/latino`, sj_race_by_block)
summary(hisp_model2)

## 
## Call:
## lm(formula = `% not completely at home pre shelter` ~ `% non hispanic/latino`, 
##     data = sj_race_by_block)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -28.6826  -2.6760  -0.0257   3.0167  16.7253 
## 
## Coefficients:
##                          Estimate Std. Error t value Pr(>|t|)    
## (Intercept)             73.005160   0.581580 125.529  < 2e-16 ***
## `% non hispanic/latino`  0.067819   0.008115   8.357 4.97e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 4.36 on 567 degrees of freedom
## Multiple R-squared:  0.1097, Adjusted R-squared:  0.1081 
## F-statistic: 69.85 on 1 and 567 DF,  p-value: 4.967e-16

Multiple regression analyses

Income, age, language, and occupants per room

# multiple regression 
modeltest <- lm(sj_ami_by_block$`% not completely at home` ~ sj_ami_by_block$`% over 125,000` + sj_age_by_block$`percent less than 30` + sj_lang_by_block$`% speaking english > well` + sj_occupants_per_room_by_block$`percent less than 1`)
summary(modeltest)

## 
## Call:
## lm(formula = sj_ami_by_block$`% not completely at home` ~ sj_ami_by_block$`% over 125,000` + 
##     sj_age_by_block$`percent less than 30` + sj_lang_by_block$`% speaking english > well` + 
##     sj_occupants_per_room_by_block$`percent less than 1`)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -23.490  -4.658  -0.797   3.973  32.270 
## 
## Coefficients:
##                                                      Estimate Std. Error
## (Intercept)                                          50.72548    4.77236
## sj_ami_by_block$`% over 125,000`                     -0.23965    0.02152
## sj_age_by_block$`percent less than 30`                0.01581    0.04209
## sj_lang_by_block$`% speaking english > well`          0.13432    0.04611
## sj_occupants_per_room_by_block$`percent less than 1` -0.02274    0.04614
##                                                      t value Pr(>|t|)    
## (Intercept)                                           10.629  < 2e-16 ***
## sj_ami_by_block$`% over 125,000`                     -11.135  < 2e-16 ***
## sj_age_by_block$`percent less than 30`                 0.376  0.70739    
## sj_lang_by_block$`% speaking english > well`           2.913  0.00372 ** 
## sj_occupants_per_room_by_block$`percent less than 1`  -0.493  0.62234    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 7.297 on 563 degrees of freedom
##   (1 observation deleted due to missingness)
## Multiple R-squared:  0.2423, Adjusted R-squared:  0.2369 
## F-statistic:    45 on 4 and 563 DF,  p-value: < 2.2e-16

Education and income

educ_income_model <- lm(sj_ami_by_block$`% not completely at home` ~ sj_ami_by_block$`% over 125,000` + sj_education_by_block$`percent associates or higher`)
summary(educ_income_model)

## 
## Call:
## lm(formula = sj_ami_by_block$`% not completely at home` ~ sj_ami_by_block$`% over 125,000` + 
##     sj_education_by_block$`percent associates or higher`)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -22.731  -4.702  -1.052   3.700  32.133 
## 
## Coefficients:
##                                                      Estimate Std. Error
## (Intercept)                                          61.47695    0.82679
## sj_ami_by_block$`% over 125,000`                     -0.15122    0.02258
## sj_education_by_block$`percent associates or higher` -0.08302    0.02219
##                                                      t value Pr(>|t|)    
## (Intercept)                                           74.356  < 2e-16 ***
## sj_ami_by_block$`% over 125,000`                      -6.697 5.15e-11 ***
## sj_education_by_block$`percent associates or higher`  -3.741 0.000202 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 7.257 on 565 degrees of freedom
##   (1 observation deleted due to missingness)
## Multiple R-squared:  0.248,  Adjusted R-squared:  0.2454 
## F-statistic: 93.18 on 2 and 565 DF,  p-value: < 2.2e-16

Internet and income

educ_income_model <- lm(sj_ami_by_block$`% not completely at home` ~ sj_ami_by_block$`% over 125,000` + sj_internet_by_block$`percent high speed`)
summary(educ_income_model)

## 
## Call:
## lm(formula = sj_ami_by_block$`% not completely at home` ~ sj_ami_by_block$`% over 125,000` + 
##     sj_internet_by_block$`percent high speed`)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -22.166  -4.531  -0.761   4.120  32.120 
## 
## Coefficients:
##                                           Estimate Std. Error t value Pr(>|t|)
## (Intercept)                               65.16907    2.28887  28.472  < 2e-16
## sj_ami_by_block$`% over 125,000`          -0.17772    0.02127  -8.356 5.03e-16
## sj_internet_by_block$`percent high speed` -0.08082    0.03393  -2.382   0.0175
##                                              
## (Intercept)                               ***
## sj_ami_by_block$`% over 125,000`          ***
## sj_internet_by_block$`percent high speed` *  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 7.309 on 565 degrees of freedom
##   (1 observation deleted due to missingness)
## Multiple R-squared:  0.2371, Adjusted R-squared:  0.2344 
## F-statistic: 87.78 on 2 and 565 DF,  p-value: < 2.2e-16

Income and Spanish language ability

income_spanish_model <- lm(sj_ami_by_block$`% not completely at home` ~ sj_ami_by_block$`% over 125,000` + sj_lang_by_block$`% not speaking spanish`)
summary(income_spanish_model)

## 
## Call:
## lm(formula = sj_ami_by_block$`% not completely at home` ~ sj_ami_by_block$`% over 125,000` + 
##     sj_lang_by_block$`% not speaking spanish`)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -23.210  -4.657  -0.882   3.991  32.085 
## 
## Coefficients:
##                                           Estimate Std. Error t value Pr(>|t|)
## (Intercept)                               62.48255    1.23594  50.555   <2e-16
## sj_ami_by_block$`% over 125,000`          -0.17861    0.02066  -8.647   <2e-16
## sj_lang_by_block$`% not speaking spanish` -0.04882    0.01961  -2.489   0.0131
##                                              
## (Intercept)                               ***
## sj_ami_by_block$`% over 125,000`          ***
## sj_lang_by_block$`% not speaking spanish` *  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 7.306 on 565 degrees of freedom
##   (1 observation deleted due to missingness)
## Multiple R-squared:  0.2378, Adjusted R-squared:  0.2351 
## F-statistic: 88.12 on 2 and 565 DF,  p-value: < 2.2e-16

This suggests that once controlling for income, Spanish language ability is no longer a strong predictor of leaving home during the shelter-in-place order.

Correlations with increase in staying home

Now I consider looking at correlations with the change in percent of devices staying completely at home since shelter-in-place started relative to the pre-shelter-in-place levels. I plot the change in percentage staying completely at home, and show linear fitting models for the change in percent staying at home, as well as the fractional increase in percent staying home.

# collect the demographic variables
sj_dem_distancing <- sj_internet_by_block %>% 
  dplyr::select(`percent high speed`, `% not completely at home`, `% Completely at Home`, blockgroup) %>% 
  left_join(sj_education_by_block %>% dplyr::select(blockgroup, `percent associates or higher`)) %>% 
  left_join(sj_ami_by_block %>% dplyr::select(blockgroup, `% over 125,000`)) %>% 
  left_join(sj_ami_by_block %>% dplyr::select(blockgroup, `% over 100,000`)) %>% 
  left_join(sj_ami_by_block %>% dplyr::select(blockgroup, `% over 75,000`)) %>% 
  left_join(sj_age_by_block %>% dplyr::select(blockgroup, `percent less than 30`)) %>% 
  left_join(sj_age_by_block %>% dplyr::select(blockgroup, `percent elderly`)) %>% 
  left_join(sj_lang_by_block %>% dplyr::select(blockgroup, `% not speaking spanish`)) %>% 
  left_join(sj_lang_by_block %>% dplyr::select(blockgroup, `% speaking english > well`)) %>% 
  left_join(sj_no_vehicles_by_block %>% dplyr::select(blockgroup, `percent with vehicles`)) %>%
  left_join(sj_occupants_per_room_by_block %>% dplyr::select(blockgroup, `percent less than 1`)) %>% 
  left_join(sj_sex_workers_by_block %>% dplyr::select(blockgroup, `% male workers`)) %>%
  left_join(sj_race_by_block %>% dplyr::select(blockgroup, `% white`, `% Asian`, `% non hispanic/latino`))

sj_dem_distancing_pre_post <- sj_dem_distancing %>% 
  left_join(sj_internet_by_block %>% dplyr::select(`% not completely at home pre shelter`, `% Completely at Home Pre Shelter`, blockgroup)) %>% 
  mutate(`% increase in staying completely home` = `% Completely at Home` - `% Completely at Home Pre Shelter`, frac_increase = `% increase in staying completely home`/`% Completely at Home Pre Shelter`)

sj_dem_distancing[is.na(sj_dem_distancing)] <- 0
sj_dem_distancing_pre_post[is.na(sj_dem_distancing_pre_post)] <- 0

saveRDS(sj_dem_distancing_pre_post, "/Users/simonespeizer/Documents/2020 Spring Quarter/CEE 218Z/covid19/Simone_Speizer/sj_socialdistancing_demdata_prepostdifs_manyvars.rds")

# sj_dem_distancing_pre_post <- readRDS("/Users/simonespeizer/Documents/2020 Spring Quarter/CEE 218Z/covid19/Simone_Speizer/sj_socialdistancing_demdata_prepostdifs_manyvars.rds")

Age

# age
sj_dem_distancing_pre_post %>%
  ggplot(aes(
  x = `percent less than 30`,
  y = `% increase in staying completely home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
labs(
    x = "Percent of residents younger than 30",
    y = "Dif in % completely at home after shelter-in-place relative to before",
    title = "San Jose: Social Distancing and Young Age Groups"
  )

young_model_dif <- lm(sj_dem_distancing_pre_post$`% increase in staying completely home` ~ sj_dem_distancing_pre_post$`percent less than 30`)
summary(young_model_dif)

## 
## Call:
## lm(formula = sj_dem_distancing_pre_post$`% increase in staying completely home` ~ 
##     sj_dem_distancing_pre_post$`percent less than 30`)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -38.099  -5.332  -0.019   5.441  30.340 
## 
## Coefficients:
##                                                   Estimate Std. Error t value
## (Intercept)                                       37.37809    1.73272  21.572
## sj_dem_distancing_pre_post$`percent less than 30` -0.29045    0.04392  -6.613
##                                                   Pr(>|t|)    
## (Intercept)                                        < 2e-16 ***
## sj_dem_distancing_pre_post$`percent less than 30` 8.72e-11 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 9.501 on 567 degrees of freedom
## Multiple R-squared:  0.0716, Adjusted R-squared:  0.06997 
## F-statistic: 43.73 on 1 and 567 DF,  p-value: 8.723e-11

young_model_frac <- lm(sj_dem_distancing_pre_post$frac_increase ~ sj_dem_distancing_pre_post$`percent less than 30`)
summary(young_model_frac)

## 
## Call:
## lm(formula = sj_dem_distancing_pre_post$frac_increase ~ sj_dem_distancing_pre_post$`percent less than 30`)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -2.06950 -0.38930 -0.09973  0.31118  2.99148 
## 
## Coefficients:
##                                                    Estimate Std. Error t value
## (Intercept)                                        2.109231   0.119117  17.707
## sj_dem_distancing_pre_post$`percent less than 30` -0.021700   0.003019  -7.187
##                                                   Pr(>|t|)    
## (Intercept)                                        < 2e-16 ***
## sj_dem_distancing_pre_post$`percent less than 30`  2.1e-12 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.6531 on 567 degrees of freedom
## Multiple R-squared:  0.08349,    Adjusted R-squared:  0.08187 
## F-statistic: 51.65 on 1 and 567 DF,  p-value: 2.102e-12

sj_dem_distancing_pre_post %>% filter(`percent elderly` < 50) %>% # get rid of extreme outliers
  ggplot(aes(
  x = `percent elderly`,
  y = `% increase in staying completely home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of residents 65 and older",
    y = "Dif in % completely at home after shelter-in-place relative to before",
    title = "San Jose: Social Distancing and Elderly Population"
  )

elderly_model_dif <- lm(`% increase in staying completely home` ~ `percent elderly`, sj_dem_distancing_pre_post %>% filter(`percent elderly` < 50))
summary(elderly_model_dif)

## 
## Call:
## lm(formula = `% increase in staying completely home` ~ `percent elderly`, 
##     data = sj_dem_distancing_pre_post %>% filter(`percent elderly` < 
##         50))
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -40.371  -5.656  -0.321   6.062  31.123 
## 
## Coefficients:
##                   Estimate Std. Error t value Pr(>|t|)    
## (Intercept)       22.25429    0.90241  24.661  < 2e-16 ***
## `percent elderly`  0.31026    0.06231   4.979 8.49e-07 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 9.636 on 564 degrees of freedom
## Multiple R-squared:  0.04211,    Adjusted R-squared:  0.04041 
## F-statistic: 24.79 on 1 and 564 DF,  p-value: 8.494e-07

elderly_model_frac <- lm(frac_increase ~ `percent elderly`, sj_dem_distancing_pre_post %>% filter(`percent elderly` < 50))
summary(elderly_model_frac)

## 
## Call:
## lm(formula = frac_increase ~ `percent elderly`, data = sj_dem_distancing_pre_post %>% 
##     filter(`percent elderly` < 50))
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -1.66307 -0.41921 -0.09938  0.33958  3.03923 
## 
## Coefficients:
##                   Estimate Std. Error t value Pr(>|t|)    
## (Intercept)       0.952358   0.062041  15.350  < 2e-16 ***
## `percent elderly` 0.025211   0.004284   5.885 6.83e-09 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.6625 on 564 degrees of freedom
## Multiple R-squared:  0.05786,    Adjusted R-squared:  0.05618 
## F-statistic: 34.63 on 1 and 564 DF,  p-value: 6.83e-09

Income

# income - less than $75000
sj_dem_distancing_pre_post %>% 
  ggplot(aes(
  x = `% over 75,000`,
  y = `% increase in staying completely home`
)) + geom_point() + 
  geom_smooth(method=lm) +
  labs(
    x = "Percent of housholds with incomes over $75,000 (50% AMI) annually",
    y = "Dif in % completely at home after shelter-in-place relative to before",
    title = "San Jose: Social Distancing and Households Above 50% AMI"
  )

income_75_model_dif <- lm(`% increase in staying completely home` ~ `% over 75,000`, sj_dem_distancing_pre_post)
summary(income_75_model_dif)

## 
## Call:
## lm(formula = `% increase in staying completely home` ~ `% over 75,000`, 
##     data = sj_dem_distancing_pre_post)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -36.384  -4.457   0.583   4.887  24.116 
## 
## Coefficients:
##                 Estimate Std. Error t value Pr(>|t|)    
## (Intercept)      8.01692    1.22405    6.55  1.3e-10 ***
## `% over 75,000`  0.29290    0.01888   15.51  < 2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 8.262 on 567 degrees of freedom
## Multiple R-squared:  0.2979, Adjusted R-squared:  0.2967 
## F-statistic: 240.6 on 1 and 567 DF,  p-value: < 2.2e-16

income_75_model_frac <- lm(frac_increase ~ `% over 75,000`, sj_dem_distancing_pre_post)
summary(income_75_model_frac)

## 
## Call:
## lm(formula = frac_increase ~ `% over 75,000`, data = sj_dem_distancing_pre_post)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -1.83204 -0.32033 -0.04946  0.27833  2.50717 
## 
## Coefficients:
##                 Estimate Std. Error t value Pr(>|t|)    
## (Intercept)     0.052042   0.085694   0.607    0.544    
## `% over 75,000` 0.019689   0.001322  14.893   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.5784 on 567 degrees of freedom
## Multiple R-squared:  0.2812, Adjusted R-squared:  0.2799 
## F-statistic: 221.8 on 1 and 567 DF,  p-value: < 2.2e-16

# income - less than $100000
sj_dem_distancing_pre_post %>% 
  ggplot(aes(
  x = `% over 100,000`,
  y = `% increase in staying completely home`
)) + geom_point() + 
  geom_smooth(method=lm) +
  labs(
    x = "Percent of housholds with incomes over $100,000 (80% AMI) annually",
    y = "Dif in % completely at home after shelter-in-place relative to before",
    title = "San Jose: Social Distancing and Households Below 80% AMI"
  )

income_100_model_dif <- lm(`% increase in staying completely home` ~ `% over 100,000`, sj_dem_distancing_pre_post)
summary(income_100_model_dif)

## 
## Call:
## lm(formula = `% increase in staying completely home` ~ `% over 100,000`, 
##     data = sj_dem_distancing_pre_post)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -36.374  -4.444   0.903   5.291  20.808 
## 
## Coefficients:
##                  Estimate Std. Error t value Pr(>|t|)    
## (Intercept)      11.28215    0.94618   11.92   <2e-16 ***
## `% over 100,000`  0.29341    0.01736   16.90   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 8.04 on 567 degrees of freedom
## Multiple R-squared:  0.3351, Adjusted R-squared:  0.3339 
## F-statistic: 285.7 on 1 and 567 DF,  p-value: < 2.2e-16

income_100_model_frac <- lm(frac_increase ~ `% over 100,000`, sj_dem_distancing_pre_post)
summary(income_100_model_frac)

## 
## Call:
## lm(formula = frac_increase ~ `% over 100,000`, data = sj_dem_distancing_pre_post)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -1.93189 -0.32427 -0.01608  0.27844  2.59102 
## 
## Coefficients:
##                  Estimate Std. Error t value Pr(>|t|)    
## (Intercept)      0.248903   0.065682    3.79 0.000167 ***
## `% over 100,000` 0.020167   0.001205   16.74  < 2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.5581 on 567 degrees of freedom
## Multiple R-squared:  0.3307, Adjusted R-squared:  0.3295 
## F-statistic: 280.1 on 1 and 567 DF,  p-value: < 2.2e-16

# income - less than $125000
sj_dem_distancing_pre_post %>% 
  ggplot(aes(
  x = `% over 125,000`,
  y = `% increase in staying completely home`
)) + geom_point() + 
  geom_smooth(method=lm) +
  labs(
    x = "Percent of housholds with incomes over $125,000 annually",
    y = "Dif in % completely at home after shelter-in-place relative to before",
    title = "San Jose: Social Distancing and Households Below $125,000"
  )

income_125_model_dif <- lm(`% increase in staying completely home` ~ `% over 125,000`, sj_dem_distancing_pre_post)
summary(income_125_model_dif)

## 
## Call:
## lm(formula = `% increase in staying completely home` ~ `% over 125,000`, 
##     data = sj_dem_distancing_pre_post)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -37.223  -4.149   0.709   5.144  21.349 
## 
## Coefficients:
##                  Estimate Std. Error t value Pr(>|t|)    
## (Intercept)      13.47554    0.78589   17.15   <2e-16 ***
## `% over 125,000`  0.31049    0.01736   17.88   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 7.884 on 567 degrees of freedom
## Multiple R-squared:  0.3606, Adjusted R-squared:  0.3595 
## F-statistic: 319.8 on 1 and 567 DF,  p-value: < 2.2e-16

income_125_model_frac <- lm(frac_increase ~ `% over 125,000`, sj_dem_distancing_pre_post)
summary(income_125_model_frac)

## 
## Call:
## lm(formula = frac_increase ~ `% over 125,000`, data = sj_dem_distancing_pre_post)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -2.01139 -0.30318  0.00149  0.27402  2.47074 
## 
## Coefficients:
##                  Estimate Std. Error t value Pr(>|t|)    
## (Intercept)      0.372176   0.053608   6.943 1.06e-11 ***
## `% over 125,000` 0.022011   0.001184  18.586  < 2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.5378 on 567 degrees of freedom
## Multiple R-squared:  0.3786, Adjusted R-squared:  0.3775 
## F-statistic: 345.4 on 1 and 567 DF,  p-value: < 2.2e-16

Language

# language
sj_dem_distancing_pre_post %>% 
  ggplot(aes(
  x = `% speaking english > well`,
  y = `% increase in staying completely home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of individuals speaking English well",
    y = "Dif in % completely at home after shelter-in-place relative to before",
    title = "San Jose: Social Distancing and English Language Ability"
  )

english_ability_model_dif <- lm(`% increase in staying completely home` ~ `% speaking english > well`, sj_dem_distancing_pre_post)
summary(english_ability_model_dif)

## 
## Call:
## lm(formula = `% increase in staying completely home` ~ `% speaking english > well`, 
##     data = sj_dem_distancing_pre_post)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -36.714  -4.743   0.360   5.067  29.631 
## 
## Coefficients:
##                             Estimate Std. Error t value Pr(>|t|)    
## (Intercept)                 -6.44360    3.75014  -1.718   0.0863 .  
## `% speaking english > well`  0.36828    0.04205   8.759   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 9.254 on 567 degrees of freedom
## Multiple R-squared:  0.1192, Adjusted R-squared:  0.1176 
## F-statistic: 76.72 on 1 and 567 DF,  p-value: < 2.2e-16

english_ability_model_frac <- lm(frac_increase ~ `% speaking english > well`, sj_dem_distancing_pre_post)
summary(english_ability_model_frac)

## 
## Call:
## lm(formula = frac_increase ~ `% speaking english > well`, data = sj_dem_distancing_pre_post)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -1.57800 -0.35263 -0.03443  0.28771  2.76709 
## 
## Coefficients:
##                              Estimate Std. Error t value Pr(>|t|)    
## (Intercept)                 -1.408245   0.252168  -5.585 3.64e-08 ***
## `% speaking english > well`  0.030260   0.002827  10.702  < 2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.6223 on 567 degrees of freedom
## Multiple R-squared:  0.1681, Adjusted R-squared:  0.1666 
## F-statistic: 114.5 on 1 and 567 DF,  p-value: < 2.2e-16

sj_dem_distancing_pre_post %>% 
  ggplot(aes(
  x = `% not speaking spanish`,
  y = `% increase in staying completely home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of individuals not speaking Spanish",
    y = "Dif in % completely at home after shelter-in-place relative to before",
    title = "San Jose: Social Distancing and Spanish Language Ability"
  )

spanish_speaking_model_dif <- lm(`% increase in staying completely home` ~ `% not speaking spanish`, sj_dem_distancing_pre_post)
summary(spanish_speaking_model_dif)

## 
## Call:
## lm(formula = `% increase in staying completely home` ~ `% not speaking spanish`, 
##     data = sj_dem_distancing_pre_post)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -40.698  -3.803   0.725   5.249  25.544 
## 
## Coefficients:
##                          Estimate Std. Error t value Pr(>|t|)    
## (Intercept)               7.60830    1.45033   5.246  2.2e-07 ***
## `% not speaking spanish`  0.23948    0.01807  13.255  < 2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 8.615 on 567 degrees of freedom
## Multiple R-squared:  0.2366, Adjusted R-squared:  0.2352 
## F-statistic: 175.7 on 1 and 567 DF,  p-value: < 2.2e-16

spanish_speaking_model_frac <- lm(frac_increase ~ `% not speaking spanish`, sj_dem_distancing_pre_post)
summary(spanish_speaking_model_frac)

## 
## Call:
## lm(formula = frac_increase ~ `% not speaking spanish`, data = sj_dem_distancing_pre_post)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -1.70690 -0.34204 -0.02644  0.27032  2.59126 
## 
## Coefficients:
##                           Estimate Std. Error t value Pr(>|t|)    
## (Intercept)              -0.038591   0.099700  -0.387    0.699    
## `% not speaking spanish`  0.016910   0.001242  13.615   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.5922 on 567 degrees of freedom
## Multiple R-squared:  0.2464, Adjusted R-squared:  0.2451 
## F-statistic: 185.4 on 1 and 567 DF,  p-value: < 2.2e-16

Occupants per room

# occupants per room
sj_dem_distancing_pre_post %>% 
  ggplot(aes(
  x = `percent less than 1`,
  y = `% increase in staying completely home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of households with 1 or fewer occupant per room",
    y = "Dif in % completely at home after shelter-in-place relative to before",
    title = "San Jose: Social Distancing and Room Occupancy"
  )

occupants_model_dif <- lm(`% increase in staying completely home` ~ `percent less than 1`, sj_dem_distancing_pre_post)
summary(occupants_model_dif)

## 
## Call:
## lm(formula = `% increase in staying completely home` ~ `percent less than 1`, 
##     data = sj_dem_distancing_pre_post)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -39.954  -4.933   0.441   5.152  27.198 
## 
## Coefficients:
##                       Estimate Std. Error t value Pr(>|t|)    
## (Intercept)           -7.01812    3.08531  -2.275   0.0233 *  
## `percent less than 1`  0.36920    0.03401  10.856   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 8.972 on 567 degrees of freedom
## Multiple R-squared:  0.1721, Adjusted R-squared:  0.1706 
## F-statistic: 117.9 on 1 and 567 DF,  p-value: < 2.2e-16

occupants_model_frac <- lm(frac_increase ~ `percent less than 1`, sj_dem_distancing_pre_post)
summary(occupants_model_frac)

## 
## Call:
## lm(formula = frac_increase ~ `percent less than 1`, data = sj_dem_distancing_pre_post)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -1.64839 -0.37296 -0.06277  0.27145  2.69847 
## 
## Coefficients:
##                        Estimate Std. Error t value Pr(>|t|)    
## (Intercept)           -1.158244   0.210792  -5.495 5.92e-08 ***
## `percent less than 1`  0.027035   0.002323  11.635  < 2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.613 on 567 degrees of freedom
## Multiple R-squared:  0.1927, Adjusted R-squared:  0.1913 
## F-statistic: 135.4 on 1 and 567 DF,  p-value: < 2.2e-16

Vehicle ownership

# vehicles - percent with no vehicles
sj_dem_distancing_pre_post %>% 
  ggplot(aes(
  x = `percent with vehicles`,
  y = `% increase in staying completely home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of housholds with vehicles available",
    y = "Dif in % completely at home after shelter-in-place relative to before",
    title = "San Jose: Social Distancing and Vehicle Availability"
  )

vehicles_model_dif <- lm(`% increase in staying completely home` ~ `percent with vehicles`, sj_dem_distancing_pre_post)
summary(vehicles_model_dif)

## 
## Call:
## lm(formula = `% increase in staying completely home` ~ `percent with vehicles`, 
##     data = sj_dem_distancing_pre_post)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -42.016  -5.916   0.106   5.629  29.764 
## 
## Coefficients:
##                         Estimate Std. Error t value Pr(>|t|)    
## (Intercept)             -9.67170    5.01754  -1.928   0.0544 .  
## `percent with vehicles`  0.37787    0.05265   7.177 2.24e-12 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 9.441 on 567 degrees of freedom
## Multiple R-squared:  0.08328,    Adjusted R-squared:  0.08166 
## F-statistic: 51.51 on 1 and 567 DF,  p-value: 2.244e-12

vehicles_model_frac <- lm(frac_increase ~ `percent with vehicles`, sj_dem_distancing_pre_post)
summary(vehicles_model_frac)

## 
## Call:
## lm(formula = frac_increase ~ `percent with vehicles`, data = sj_dem_distancing_pre_post)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -1.8015 -0.4422 -0.1204  0.3100  2.8799 
## 
## Coefficients:
##                          Estimate Std. Error t value Pr(>|t|)    
## (Intercept)             -0.998778   0.349695  -2.856  0.00445 ** 
## `percent with vehicles`  0.023946   0.003669   6.526  1.5e-10 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.658 on 567 degrees of freedom
## Multiple R-squared:  0.06986,    Adjusted R-squared:  0.06822 
## F-statistic: 42.58 on 1 and 567 DF,  p-value: 1.503e-10

Education

sj_dem_distancing_pre_post %>% 
  ggplot(aes(
  x = `percent associates or higher`,
  y = `% increase in staying completely home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of people with an degree at Associate's level or higher",
    y = "Dif in % completely at home after shelter-in-place relative to before",
    title = "San Jose: Social Distancing and Education"
  )

educ_model_dif <- lm(`% increase in staying completely home` ~ `percent associates or higher`, sj_dem_distancing_pre_post)
summary(educ_model_dif)

## 
## Call:
## lm(formula = `% increase in staying completely home` ~ `percent associates or higher`, 
##     data = sj_dem_distancing_pre_post)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -40.284  -3.059   1.107   5.124  22.190 
## 
## Coefficients:
##                                Estimate Std. Error t value Pr(>|t|)    
## (Intercept)                    13.14752    0.91413   14.38   <2e-16 ***
## `percent associates or higher`  0.27737    0.01794   15.46   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 8.27 on 567 degrees of freedom
## Multiple R-squared:  0.2966, Adjusted R-squared:  0.2954 
## F-statistic: 239.1 on 1 and 567 DF,  p-value: < 2.2e-16

educ_model_frac <- lm(frac_increase ~ `percent associates or higher`, sj_dem_distancing_pre_post)
summary(educ_model_frac)

## 
## Call:
## lm(formula = frac_increase ~ `percent associates or higher`, 
##     data = sj_dem_distancing_pre_post)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -2.26600 -0.31586  0.00808  0.28974  2.48982 
## 
## Coefficients:
##                                Estimate Std. Error t value Pr(>|t|)    
## (Intercept)                    0.357339   0.062837   5.687 2.07e-08 ***
## `percent associates or higher` 0.019484   0.001233  15.802  < 2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.5685 on 567 degrees of freedom
## Multiple R-squared:  0.3057, Adjusted R-squared:  0.3045 
## F-statistic: 249.7 on 1 and 567 DF,  p-value: < 2.2e-16

Internet

sj_dem_distancing_pre_post %>% 
  ggplot(aes(
  x = `percent high speed`,
  y = `% increase in staying completely home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of households with broadband such as cable, fiber optic or DSL",
    y = "Dif in % completely at home after shelter-in-place relative to before",
    title = "San Jose: Social Distancing and High Speed Internet"
  )

internet_model_dif <- lm(`% increase in staying completely home` ~ `percent high speed`, sj_dem_distancing_pre_post)
summary(internet_model_dif)

## 
## Call:
## lm(formula = `% increase in staying completely home` ~ `percent high speed`, 
##     data = sj_dem_distancing_pre_post)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -39.462  -4.781   0.411   5.399  26.685 
## 
## Coefficients:
##                      Estimate Std. Error t value Pr(>|t|)    
## (Intercept)          -3.54147    2.41478  -1.467    0.143    
## `percent high speed`  0.36963    0.02964  12.472   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 8.735 on 567 degrees of freedom
## Multiple R-squared:  0.2153, Adjusted R-squared:  0.2139 
## F-statistic: 155.5 on 1 and 567 DF,  p-value: < 2.2e-16

internet_model_frac <- lm(frac_increase ~ `percent high speed`, sj_dem_distancing_pre_post)
summary(internet_model_frac)

## 
## Call:
## lm(formula = frac_increase ~ `percent high speed`, data = sj_dem_distancing_pre_post)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -1.7085 -0.3548 -0.0774  0.2784  2.6841 
## 
## Coefficients:
##                       Estimate Std. Error t value Pr(>|t|)    
## (Intercept)          -0.677570   0.169360  -4.001 7.15e-05 ***
## `percent high speed`  0.024258   0.002079  11.671  < 2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.6126 on 567 degrees of freedom
## Multiple R-squared:  0.1937, Adjusted R-squared:  0.1923 
## F-statistic: 136.2 on 1 and 567 DF,  p-value: < 2.2e-16

Sex of workers

sj_dem_distancing_pre_post %>% 
  ggplot(aes(
  x = `% male workers`,
  y = `% increase in staying completely home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of workers that are male",
    y = "Dif in % completely at home after shelter-in-place relative to before",
    title = "San Jose: Social Distancing and Sex of Workers"
  )

sex_workers_model_dif <- lm(`% increase in staying completely home` ~ `% male workers`, sj_dem_distancing_pre_post)
summary(sex_workers_model_dif)

## 
## Call:
## lm(formula = `% increase in staying completely home` ~ `% male workers`, 
##     data = sj_dem_distancing_pre_post)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -42.390  -6.247  -0.230   6.165  30.825 
## 
## Coefficients:
##                  Estimate Std. Error t value Pr(>|t|)    
## (Intercept)      24.14743    6.41049   3.767 0.000183 ***
## `% male workers`  0.03886    0.11956   0.325 0.745277    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 9.859 on 567 degrees of freedom
## Multiple R-squared:  0.0001863,  Adjusted R-squared:  -0.001577 
## F-statistic: 0.1056 on 1 and 567 DF,  p-value: 0.7453

sex_workers_model_frac <- lm(frac_increase ~ `% male workers`, sj_dem_distancing_pre_post)
summary(sex_workers_model_frac)

## 
## Call:
## lm(formula = frac_increase ~ `% male workers`, data = sj_dem_distancing_pre_post)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -1.8587 -0.4506 -0.1051  0.3343  2.8996 
## 
## Coefficients:
##                  Estimate Std. Error t value Pr(>|t|)
## (Intercept)      0.715952   0.442955   1.616    0.107
## `% male workers` 0.010469   0.008261   1.267    0.206
## 
## Residual standard error: 0.6813 on 567 degrees of freedom
## Multiple R-squared:  0.002824,   Adjusted R-squared:  0.001065 
## F-statistic: 1.606 on 1 and 567 DF,  p-value: 0.2056

Race

# white
sj_dem_distancing_pre_post %>% 
  ggplot(aes(
  x = `% white`,
  y = `% increase in staying completely home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of residents that are white",
    y = "Dif in % completely at home after shelter-in-place relative to before",
    title = "San Jose: Social Distancing and White Residents"
  )

white_model_dif <- lm(`% increase in staying completely home` ~ `% white`, sj_dem_distancing_pre_post)
summary(white_model_dif)

## 
## Call:
## lm(formula = `% increase in staying completely home` ~ `% white`, 
##     data = sj_dem_distancing_pre_post)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -42.164  -5.746  -0.222   6.115  31.833 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)    
## (Intercept) 24.70515    0.91587   26.98   <2e-16 ***
## `% white`    0.03534    0.01900    1.86   0.0634 .  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 9.83 on 567 degrees of freedom
## Multiple R-squared:  0.006066,   Adjusted R-squared:  0.004314 
## F-statistic: 3.461 on 1 and 567 DF,  p-value: 0.06336

white_model_frac <- lm(frac_increase ~ `% white`, sj_dem_distancing_pre_post)
summary(white_model_frac)

## 
## Call:
## lm(formula = frac_increase ~ `% white`, data = sj_dem_distancing_pre_post)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -1.8034 -0.4237 -0.0745  0.3101  2.9611 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)    
## (Intercept) 0.995999   0.062182  16.017  < 2e-16 ***
## `% white`   0.006506   0.001290   5.044 6.15e-07 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.6674 on 567 degrees of freedom
## Multiple R-squared:  0.04294,    Adjusted R-squared:  0.04126 
## F-statistic: 25.44 on 1 and 567 DF,  p-value: 6.146e-07

# asian
sj_dem_distancing_pre_post %>% 
  ggplot(aes(
  x = `% Asian`,
  y = `% increase in staying completely home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of residents that are Asian",
    y = "Dif in % completely at home after shelter-in-place relative to before",
    title = "San Jose: Social Distancing and Asian Residents"
  )

asian_model_dif <- lm(`% increase in staying completely home` ~ `% Asian`, sj_dem_distancing_pre_post)
summary(asian_model_dif)

## 
## Call:
## lm(formula = `% increase in staying completely home` ~ `% Asian`, 
##     data = sj_dem_distancing_pre_post)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -42.123  -5.662  -0.366   6.008  26.118 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)    
## (Intercept) 21.86541    0.71307  30.664  < 2e-16 ***
## `% Asian`    0.13438    0.01829   7.346 7.13e-13 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 9.422 on 567 degrees of freedom
## Multiple R-squared:  0.08691,    Adjusted R-squared:  0.0853 
## F-statistic: 53.97 on 1 and 567 DF,  p-value: 7.132e-13

asian_model_frac <- lm(frac_increase ~ `% Asian`, sj_dem_distancing_pre_post)
summary(asian_model_frac)

## 
## Call:
## lm(formula = frac_increase ~ `% Asian`, data = sj_dem_distancing_pre_post)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -1.8110 -0.4508 -0.1314  0.3053  2.9293 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)    
## (Intercept) 1.114130   0.050981  21.854  < 2e-16 ***
## `% Asian`   0.004990   0.001308   3.816 0.000151 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.6736 on 567 degrees of freedom
## Multiple R-squared:  0.02504,    Adjusted R-squared:  0.02332 
## F-statistic: 14.56 on 1 and 567 DF,  p-value: 0.0001505

# hispanic/latino
sj_dem_distancing_pre_post %>% 
  ggplot(aes(
  x = `% non hispanic/latino`,
  y = `% increase in staying completely home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of residents that are not Hispanic or Latino",
    y = "Dif in % completely at home after shelter-in-place relative to before",
    title = "San Jose: Social Distancing and Hispanic/Latino Residents"
  )

hisp_model_dif <- lm(`% increase in staying completely home` ~ `% non hispanic/latino`, sj_dem_distancing_pre_post)
summary(hisp_model_dif)

## 
## Call:
## lm(formula = `% increase in staying completely home` ~ `% non hispanic/latino`, 
##     data = sj_dem_distancing_pre_post)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -41.536  -3.920   0.957   5.121  23.810 
## 
## Coefficients:
##                         Estimate Std. Error t value Pr(>|t|)    
## (Intercept)             10.78733    1.12400   9.597   <2e-16 ***
## `% non hispanic/latino`  0.22693    0.01568  14.469   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 8.427 on 567 degrees of freedom
## Multiple R-squared:  0.2697, Adjusted R-squared:  0.2684 
## F-statistic: 209.4 on 1 and 567 DF,  p-value: < 2.2e-16

hisp_model_frac <- lm(frac_increase ~ `% non hispanic/latino`, sj_dem_distancing_pre_post)
summary(hisp_model_frac)

## 
## Call:
## lm(formula = frac_increase ~ `% non hispanic/latino`, data = sj_dem_distancing_pre_post)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -1.7671 -0.3543 -0.0187  0.2943  2.5558 
## 
## Coefficients:
##                         Estimate Std. Error t value Pr(>|t|)    
## (Intercept)             0.209446   0.077812   2.692  0.00732 ** 
## `% non hispanic/latino` 0.015678   0.001086  14.440  < 2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.5834 on 567 degrees of freedom
## Multiple R-squared:  0.2689, Adjusted R-squared:  0.2676 
## F-statistic: 208.5 on 1 and 567 DF,  p-value: < 2.2e-16

Multiple regression for increases in staying at home

Multiple regression analysis: income, internet, Spanish language ability, and occupants per room.

First with difference in percents.

difs_model <- lm(sj_dem_distancing_pre_post$`% increase in staying completely home` ~ sj_dem_distancing_pre_post$`% over 125,000` + sj_dem_distancing_pre_post$`% not speaking spanish` + sj_dem_distancing_pre_post$`percent less than 1` + sj_dem_distancing_pre_post$`percent high speed`)
summary(difs_model)

## 
## Call:
## lm(formula = sj_dem_distancing_pre_post$`% increase in staying completely home` ~ 
##     sj_dem_distancing_pre_post$`% over 125,000` + sj_dem_distancing_pre_post$`% not speaking spanish` + 
##         sj_dem_distancing_pre_post$`percent less than 1` + sj_dem_distancing_pre_post$`percent high speed`)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -37.172  -3.947   0.889   5.005  20.897 
## 
## Coefficients:
##                                                      Estimate Std. Error
## (Intercept)                                          5.888624   3.296538
## sj_dem_distancing_pre_post$`% over 125,000`          0.233541   0.024654
## sj_dem_distancing_pre_post$`% not speaking spanish`  0.080083   0.025519
## sj_dem_distancing_pre_post$`percent less than 1`    -0.006499   0.043997
## sj_dem_distancing_pre_post$`percent high speed`      0.063404   0.036865
##                                                     t value Pr(>|t|)    
## (Intercept)                                           1.786  0.07459 .  
## sj_dem_distancing_pre_post$`% over 125,000`           9.473  < 2e-16 ***
## sj_dem_distancing_pre_post$`% not speaking spanish`   3.138  0.00179 ** 
## sj_dem_distancing_pre_post$`percent less than 1`     -0.148  0.88262    
## sj_dem_distancing_pre_post$`percent high speed`       1.720  0.08599 .  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 7.758 on 564 degrees of freedom
## Multiple R-squared:  0.3843, Adjusted R-squared:  0.3799 
## F-statistic: 87.99 on 4 and 564 DF,  p-value: < 2.2e-16

Second with fractional change.

frac_model <- lm(sj_dem_distancing_pre_post$frac_increase ~ sj_dem_distancing_pre_post$`% over 125,000` + sj_dem_distancing_pre_post$`% not speaking spanish` + sj_dem_distancing_pre_post$`percent less than 1` + sj_dem_distancing_pre_post$`percent high speed`)
summary(frac_model)

## 
## Call:
## lm(formula = sj_dem_distancing_pre_post$frac_increase ~ sj_dem_distancing_pre_post$`% over 125,000` + 
##     sj_dem_distancing_pre_post$`% not speaking spanish` + sj_dem_distancing_pre_post$`percent less than 1` + 
##     sj_dem_distancing_pre_post$`percent high speed`)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -1.98845 -0.30220  0.01756  0.26994  2.37443 
## 
## Coefficients:
##                                                       Estimate Std. Error
## (Intercept)                                         -0.1122792  0.2251531
## sj_dem_distancing_pre_post$`% over 125,000`          0.0174733  0.0016838
## sj_dem_distancing_pre_post$`% not speaking spanish`  0.0054210  0.0017429
## sj_dem_distancing_pre_post$`percent less than 1`     0.0023581  0.0030050
## sj_dem_distancing_pre_post$`percent high speed`      0.0004595  0.0025178
##                                                     t value Pr(>|t|)    
## (Intercept)                                          -0.499  0.61820    
## sj_dem_distancing_pre_post$`% over 125,000`          10.377  < 2e-16 ***
## sj_dem_distancing_pre_post$`% not speaking spanish`   3.110  0.00196 ** 
## sj_dem_distancing_pre_post$`percent less than 1`      0.785  0.43295    
## sj_dem_distancing_pre_post$`percent high speed`       0.183  0.85525    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.5299 on 564 degrees of freedom
## Multiple R-squared:    0.4,  Adjusted R-squared:  0.3957 
## F-statistic:    94 on 4 and 564 DF,  p-value: < 2.2e-16

Multiple regression analysis: income and Spanish language ability

difs_model_inc_span <- lm(sj_dem_distancing_pre_post$`% increase in staying completely home` ~ sj_dem_distancing_pre_post$`% over 125,000` + sj_dem_distancing_pre_post$`% not speaking spanish`)
summary(difs_model_inc_span)

## 
## Call:
## lm(formula = sj_dem_distancing_pre_post$`% increase in staying completely home` ~ 
##     sj_dem_distancing_pre_post$`% over 125,000` + sj_dem_distancing_pre_post$`% not speaking spanish`)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -37.056  -4.101   1.030   5.118  21.602 
## 
## Coefficients:
##                                                     Estimate Std. Error t value
## (Intercept)                                          8.90321    1.31193   6.786
## sj_dem_distancing_pre_post$`% over 125,000`          0.25153    0.02189  11.493
## sj_dem_distancing_pre_post$`% not speaking spanish`  0.08996    0.02084   4.316
##                                                     Pr(>|t|)    
## (Intercept)                                         2.91e-11 ***
## sj_dem_distancing_pre_post$`% over 125,000`          < 2e-16 ***
## sj_dem_distancing_pre_post$`% not speaking spanish` 1.87e-05 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 7.764 on 566 degrees of freedom
## Multiple R-squared:  0.381,  Adjusted R-squared:  0.3788 
## F-statistic: 174.2 on 2 and 566 DF,  p-value: < 2.2e-16

frac_model_inc_span <- lm(sj_dem_distancing_pre_post$frac_increase ~ sj_dem_distancing_pre_post$`% over 125,000` + sj_dem_distancing_pre_post$`% not speaking spanish`)
summary(frac_model_inc_span)

## 
## Call:
## lm(formula = sj_dem_distancing_pre_post$frac_increase ~ sj_dem_distancing_pre_post$`% over 125,000` + 
##     sj_dem_distancing_pre_post$`% not speaking spanish`)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -1.99970 -0.29963  0.02114  0.26759  2.38885 
## 
## Coefficients:
##                                                     Estimate Std. Error t value
## (Intercept)                                         0.053571   0.089426   0.599
## sj_dem_distancing_pre_post$`% over 125,000`         0.017902   0.001492  12.000
## sj_dem_distancing_pre_post$`% not speaking spanish` 0.006268   0.001421   4.412
##                                                     Pr(>|t|)    
## (Intercept)                                            0.549    
## sj_dem_distancing_pre_post$`% over 125,000`          < 2e-16 ***
## sj_dem_distancing_pre_post$`% not speaking spanish` 1.22e-05 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.5293 on 566 degrees of freedom
## Multiple R-squared:  0.3992, Adjusted R-squared:  0.3971 
## F-statistic: 188.1 on 2 and 566 DF,  p-value: < 2.2e-16

When only accounting for for income, Spanish language ability is only slightly relevant, though the result is still nontrivial. Let’s try accounting for both education and income level.

Multiple regression analysis: income, education, and Spanish language ability

difs_model_inc_span_educ <- lm(sj_dem_distancing_pre_post$`% increase in staying completely home` ~ sj_dem_distancing_pre_post$`% over 125,000` + sj_dem_distancing_pre_post$`% not speaking spanish` +  sj_dem_distancing_pre_post$`percent associates or higher`)
summary(difs_model_inc_span_educ)

## 
## Call:
## lm(formula = sj_dem_distancing_pre_post$`% increase in staying completely home` ~ 
##     sj_dem_distancing_pre_post$`% over 125,000` + sj_dem_distancing_pre_post$`% not speaking spanish` + 
##         sj_dem_distancing_pre_post$`percent associates or higher`)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -36.754  -3.631   1.019   5.037  21.054 
## 
## Coefficients:
##                                                           Estimate Std. Error
## (Intercept)                                                9.91819    1.34021
## sj_dem_distancing_pre_post$`% over 125,000`                0.21759    0.02420
## sj_dem_distancing_pre_post$`% not speaking spanish`        0.03620    0.02672
## sj_dem_distancing_pre_post$`percent associates or higher`  0.09667    0.03045
##                                                           t value Pr(>|t|)    
## (Intercept)                                                 7.400 4.94e-13 ***
## sj_dem_distancing_pre_post$`% over 125,000`                 8.991  < 2e-16 ***
## sj_dem_distancing_pre_post$`% not speaking spanish`         1.355  0.17609    
## sj_dem_distancing_pre_post$`percent associates or higher`   3.175  0.00158 ** 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 7.703 on 565 degrees of freedom
## Multiple R-squared:  0.3919, Adjusted R-squared:  0.3886 
## F-statistic: 121.4 on 3 and 565 DF,  p-value: < 2.2e-16

frac_model_inc_span_educ <- lm(sj_dem_distancing_pre_post$frac_increase ~ sj_dem_distancing_pre_post$`% over 125,000` + sj_dem_distancing_pre_post$`% not speaking spanish` +  sj_dem_distancing_pre_post$`percent associates or higher`)
summary(frac_model_inc_span_educ)

## 
## Call:
## lm(formula = sj_dem_distancing_pre_post$frac_increase ~ sj_dem_distancing_pre_post$`% over 125,000` + 
##     sj_dem_distancing_pre_post$`% not speaking spanish` + sj_dem_distancing_pre_post$`percent associates or higher`)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -1.94691 -0.29298  0.00853  0.26299  2.41914 
## 
## Coefficients:
##                                                           Estimate Std. Error
## (Intercept)                                               0.120677   0.091402
## sj_dem_distancing_pre_post$`% over 125,000`               0.015658   0.001650
## sj_dem_distancing_pre_post$`% not speaking spanish`       0.002714   0.001823
## sj_dem_distancing_pre_post$`percent associates or higher` 0.006391   0.002076
##                                                           t value Pr(>|t|)    
## (Intercept)                                                 1.320  0.18727    
## sj_dem_distancing_pre_post$`% over 125,000`                 9.487  < 2e-16 ***
## sj_dem_distancing_pre_post$`% not speaking spanish`         1.489  0.13700    
## sj_dem_distancing_pre_post$`percent associates or higher`   3.078  0.00218 ** 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.5253 on 565 degrees of freedom
## Multiple R-squared:  0.4092, Adjusted R-squared:  0.406 
## F-statistic: 130.4 on 3 and 565 DF,  p-value: < 2.2e-16

The effect of Spanish language speaking vanishes when accounting for both education and income.

Multiple regression analysis: income, English language ability and education

difs_model_inc_eng_educ <- lm(sj_dem_distancing_pre_post$`% increase in staying completely home` ~ sj_dem_distancing_pre_post$`% over 125,000` + sj_dem_distancing_pre_post$`% speaking english > well` +  sj_dem_distancing_pre_post$`percent associates or higher`)
summary(difs_model_inc_eng_educ)

## 
## Call:
## lm(formula = sj_dem_distancing_pre_post$`% increase in staying completely home` ~ 
##     sj_dem_distancing_pre_post$`% over 125,000` + sj_dem_distancing_pre_post$`% speaking english > well` + 
##         sj_dem_distancing_pre_post$`percent associates or higher`)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -38.655  -3.721   1.045   4.889  21.069 
## 
## Coefficients:
##                                                           Estimate Std. Error
## (Intercept)                                               24.41313    3.70697
## sj_dem_distancing_pre_post$`% over 125,000`                0.23917    0.02411
## sj_dem_distancing_pre_post$`% speaking english > well`    -0.18226    0.05007
## sj_dem_distancing_pre_post$`percent associates or higher`  0.17304    0.02710
##                                                           t value Pr(>|t|)    
## (Intercept)                                                 6.586 1.04e-10 ***
## sj_dem_distancing_pre_post$`% over 125,000`                 9.920  < 2e-16 ***
## sj_dem_distancing_pre_post$`% speaking english > well`     -3.640 0.000298 ***
## sj_dem_distancing_pre_post$`percent associates or higher`   6.385 3.58e-10 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 7.627 on 565 degrees of freedom
## Multiple R-squared:  0.4039, Adjusted R-squared:  0.4007 
## F-statistic: 127.6 on 3 and 565 DF,  p-value: < 2.2e-16

frac_model_inc_eng_educ <- lm(sj_dem_distancing_pre_post$frac_increase ~ sj_dem_distancing_pre_post$`% over 125,000` + sj_dem_distancing_pre_post$`% speaking english > well` +  sj_dem_distancing_pre_post$`percent associates or higher`)
summary(frac_model_inc_eng_educ)

## 
## Call:
## lm(formula = sj_dem_distancing_pre_post$frac_increase ~ sj_dem_distancing_pre_post$`% over 125,000` + 
##     sj_dem_distancing_pre_post$`% speaking english > well` + 
##     sj_dem_distancing_pre_post$`percent associates or higher`)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -1.93778 -0.29842  0.01177  0.26564  2.43658 
## 
## Coefficients:
##                                                            Estimate Std. Error
## (Intercept)                                                0.497226   0.255573
## sj_dem_distancing_pre_post$`% over 125,000`                0.016380   0.001662
## sj_dem_distancing_pre_post$`% speaking english > well`    -0.003798   0.003452
## sj_dem_distancing_pre_post$`percent associates or higher`  0.009397   0.001868
##                                                           t value Pr(>|t|)    
## (Intercept)                                                 1.946   0.0522 .  
## sj_dem_distancing_pre_post$`% over 125,000`                 9.854  < 2e-16 ***
## sj_dem_distancing_pre_post$`% speaking english > well`     -1.100   0.2717    
## sj_dem_distancing_pre_post$`percent associates or higher`   5.029 6.62e-07 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.5258 on 565 degrees of freedom
## Multiple R-squared:  0.4081, Adjusted R-squared:  0.405 
## F-statistic: 129.9 on 3 and 565 DF,  p-value: < 2.2e-16

English language ability is actually a slightly better predictor than Spanish language ability, when also accounting for education and income.

Multiple regression analysis: income, English language ability, education, Spanish language ability, and vehicle ownership

difs_model_lots <- lm(sj_dem_distancing_pre_post$`% increase in staying completely home` ~ sj_dem_distancing_pre_post$`% over 125,000` + sj_dem_distancing_pre_post$`% speaking english > well` +  sj_dem_distancing_pre_post$`percent associates or higher` + sj_dem_distancing_pre_post$`% not speaking spanish` + sj_dem_distancing_pre_post$`percent with vehicles`)
summary(difs_model_lots)

## 
## Call:
## lm(formula = sj_dem_distancing_pre_post$`% increase in staying completely home` ~ 
##     sj_dem_distancing_pre_post$`% over 125,000` + sj_dem_distancing_pre_post$`% speaking english > well` + 
##         sj_dem_distancing_pre_post$`percent associates or higher` + 
##         sj_dem_distancing_pre_post$`% not speaking spanish` + 
##         sj_dem_distancing_pre_post$`percent with vehicles`)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -39.193  -3.558   1.047   4.884  21.144 
## 
## Coefficients:
##                                                           Estimate Std. Error
## (Intercept)                                               15.31547    5.48658
## sj_dem_distancing_pre_post$`% over 125,000`                0.21569    0.02589
## sj_dem_distancing_pre_post$`% speaking english > well`    -0.19520    0.05013
## sj_dem_distancing_pre_post$`percent associates or higher`  0.14682    0.03251
## sj_dem_distancing_pre_post$`% not speaking spanish`        0.04655    0.02647
## sj_dem_distancing_pre_post$`percent with vehicles`         0.09292    0.04703
##                                                           t value Pr(>|t|)    
## (Intercept)                                                 2.791  0.00543 ** 
## sj_dem_distancing_pre_post$`% over 125,000`                 8.330 6.18e-16 ***
## sj_dem_distancing_pre_post$`% speaking english > well`     -3.894  0.00011 ***
## sj_dem_distancing_pre_post$`percent associates or higher`   4.516 7.68e-06 ***
## sj_dem_distancing_pre_post$`% not speaking spanish`         1.759  0.07911 .  
## sj_dem_distancing_pre_post$`percent with vehicles`          1.976  0.04869 *  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 7.595 on 563 degrees of freedom
## Multiple R-squared:  0.4108, Adjusted R-squared:  0.4056 
## F-statistic: 78.51 on 5 and 563 DF,  p-value: < 2.2e-16

frac_model_lots <- lm(sj_dem_distancing_pre_post$frac_increase ~ sj_dem_distancing_pre_post$`% over 125,000` + sj_dem_distancing_pre_post$`% speaking english > well` +  sj_dem_distancing_pre_post$`percent associates or higher` + sj_dem_distancing_pre_post$`% not speaking spanish` + sj_dem_distancing_pre_post$`percent with vehicles`)
summary(frac_model_lots)

## 
## Call:
## lm(formula = sj_dem_distancing_pre_post$frac_increase ~ sj_dem_distancing_pre_post$`% over 125,000` + 
##     sj_dem_distancing_pre_post$`% speaking english > well` + 
##     sj_dem_distancing_pre_post$`percent associates or higher` + 
##     sj_dem_distancing_pre_post$`% not speaking spanish` + sj_dem_distancing_pre_post$`percent with vehicles`)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -1.94356 -0.30435  0.00216  0.26876  2.41224 
## 
## Coefficients:
##                                                            Estimate Std. Error
## (Intercept)                                                0.208851   0.379462
## sj_dem_distancing_pre_post$`% over 125,000`                0.015542   0.001791
## sj_dem_distancing_pre_post$`% speaking english > well`    -0.004389   0.003467
## sj_dem_distancing_pre_post$`percent associates or higher`  0.007527   0.002248
## sj_dem_distancing_pre_post$`% not speaking spanish`        0.002957   0.001830
## sj_dem_distancing_pre_post$`percent with vehicles`         0.002458   0.003253
##                                                           t value Pr(>|t|)    
## (Intercept)                                                 0.550  0.58227    
## sj_dem_distancing_pre_post$`% over 125,000`                 8.679  < 2e-16 ***
## sj_dem_distancing_pre_post$`% speaking english > well`     -1.266  0.20610    
## sj_dem_distancing_pre_post$`percent associates or higher`   3.348  0.00087 ***
## sj_dem_distancing_pre_post$`% not speaking spanish`         1.615  0.10677    
## sj_dem_distancing_pre_post$`percent with vehicles`          0.756  0.45020    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.5253 on 563 degrees of freedom
## Multiple R-squared:  0.4113, Adjusted R-squared:  0.4061 
## F-statistic: 78.67 on 5 and 563 DF,  p-value: < 2.2e-16

The main important variables are education and income, with potentially some effect of English language ability.

Multiple regression analysis: income and education

difs_model_inc_educ <- lm(sj_dem_distancing_pre_post$`% increase in staying completely home` ~ sj_dem_distancing_pre_post$`% over 125,000` +  sj_dem_distancing_pre_post$`percent associates or higher`)
summary(difs_model_inc_educ)

## 
## Call:
## lm(formula = sj_dem_distancing_pre_post$`% increase in staying completely home` ~ 
##     sj_dem_distancing_pre_post$`% over 125,000` + sj_dem_distancing_pre_post$`percent associates or higher`)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -36.671  -3.735   1.128   4.928  20.844 
## 
## Coefficients:
##                                                           Estimate Std. Error
## (Intercept)                                               11.29393    0.87510
## sj_dem_distancing_pre_post$`% over 125,000`                0.22262    0.02393
## sj_dem_distancing_pre_post$`percent associates or higher`  0.12280    0.02357
##                                                           t value Pr(>|t|)    
## (Intercept)                                                12.906  < 2e-16 ***
## sj_dem_distancing_pre_post$`% over 125,000`                 9.302  < 2e-16 ***
## sj_dem_distancing_pre_post$`percent associates or higher`   5.209 2.66e-07 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 7.709 on 566 degrees of freedom
## Multiple R-squared:  0.3899, Adjusted R-squared:  0.3877 
## F-statistic: 180.9 on 2 and 566 DF,  p-value: < 2.2e-16

frac_model_inc_educ <- lm(sj_dem_distancing_pre_post$`frac_increase` ~ sj_dem_distancing_pre_post$`% over 125,000` +  sj_dem_distancing_pre_post$`percent associates or higher`)
summary(frac_model_inc_educ)

## 
## Call:
## lm(formula = sj_dem_distancing_pre_post$frac_increase ~ sj_dem_distancing_pre_post$`% over 125,000` + 
##     sj_dem_distancing_pre_post$`percent associates or higher`)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -1.9338 -0.2932  0.0080  0.2672  2.4497 
## 
## Coefficients:
##                                                           Estimate Std. Error
## (Intercept)                                               0.223825   0.059702
## sj_dem_distancing_pre_post$`% over 125,000`               0.016035   0.001633
## sj_dem_distancing_pre_post$`percent associates or higher` 0.008350   0.001608
##                                                           t value Pr(>|t|)    
## (Intercept)                                                 3.749 0.000196 ***
## sj_dem_distancing_pre_post$`% over 125,000`                 9.821  < 2e-16 ***
## sj_dem_distancing_pre_post$`percent associates or higher`   5.192  2.9e-07 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.5259 on 566 degrees of freedom
## Multiple R-squared:  0.4068, Adjusted R-squared:  0.4047 
## F-statistic: 194.1 on 2 and 566 DF,  p-value: < 2.2e-16

Comparing this to earlier models, we see that adding the English language ability variable does add some predictive power, though not much, and adding the vehicle ownership and Spanish language ability variables have negligible effects.

We now consider adding race into the regressions.

Multiple regression analysis: Hispanic/Latino and income

difs_model_inc_hisp <- lm(sj_dem_distancing_pre_post$`% increase in staying completely home` ~ sj_dem_distancing_pre_post$`% over 125,000` +  sj_dem_distancing_pre_post$`% non hispanic/latino`)
summary(difs_model_inc_hisp)

## 
## Call:
## lm(formula = sj_dem_distancing_pre_post$`% increase in staying completely home` ~ 
##     sj_dem_distancing_pre_post$`% over 125,000` + sj_dem_distancing_pre_post$`% non hispanic/latino`)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -37.105  -3.779   1.002   5.099  20.657 
## 
## Coefficients:
##                                                    Estimate Std. Error t value
## (Intercept)                                         9.79763    1.03125   9.501
## sj_dem_distancing_pre_post$`% over 125,000`         0.23490    0.02209  10.635
## sj_dem_distancing_pre_post$`% non hispanic/latino`  0.09968    0.01867   5.340
##                                                    Pr(>|t|)    
## (Intercept)                                         < 2e-16 ***
## sj_dem_distancing_pre_post$`% over 125,000`         < 2e-16 ***
## sj_dem_distancing_pre_post$`% non hispanic/latino` 1.35e-07 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 7.7 on 566 degrees of freedom
## Multiple R-squared:  0.3913, Adjusted R-squared:  0.3892 
## F-statistic: 181.9 on 2 and 566 DF,  p-value: < 2.2e-16

frac_model_inc_hisp <- lm(sj_dem_distancing_pre_post$frac_increase ~ sj_dem_distancing_pre_post$`% over 125,000` +  sj_dem_distancing_pre_post$`% non hispanic/latino`)
summary(frac_model_inc_hisp)

## 
## Call:
## lm(formula = sj_dem_distancing_pre_post$frac_increase ~ sj_dem_distancing_pre_post$`% over 125,000` + 
##     sj_dem_distancing_pre_post$`% non hispanic/latino`)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -2.0007 -0.3147  0.0192  0.2745  2.4204 
## 
## Coefficients:
##                                                    Estimate Std. Error t value
## (Intercept)                                        0.137068   0.070560   1.943
## sj_dem_distancing_pre_post$`% over 125,000`        0.017179   0.001511  11.367
## sj_dem_distancing_pre_post$`% non hispanic/latino` 0.006372   0.001277   4.989
##                                                    Pr(>|t|)    
## (Intercept)                                          0.0526 .  
## sj_dem_distancing_pre_post$`% over 125,000`         < 2e-16 ***
## sj_dem_distancing_pre_post$`% non hispanic/latino`  8.1e-07 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.5268 on 566 degrees of freedom
## Multiple R-squared:  0.4048, Adjusted R-squared:  0.4026 
## F-statistic: 192.4 on 2 and 566 DF,  p-value: < 2.2e-16

Multiple regression analysis: Hispanic/Latino, income, and education

difs_model_inc_hisp_educ <- lm(sj_dem_distancing_pre_post$`% increase in staying completely home` ~ sj_dem_distancing_pre_post$`% over 125,000` +  sj_dem_distancing_pre_post$`% non hispanic/latino` + sj_dem_distancing_pre_post$`percent associates or higher`)
summary(difs_model_inc_hisp_educ)

## 
## Call:
## lm(formula = sj_dem_distancing_pre_post$`% increase in staying completely home` ~ 
##     sj_dem_distancing_pre_post$`% over 125,000` + sj_dem_distancing_pre_post$`% non hispanic/latino` + 
##         sj_dem_distancing_pre_post$`percent associates or higher`)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -37.096  -3.705   1.041   5.109  20.632 
## 
## Coefficients:
##                                                           Estimate Std. Error
## (Intercept)                                                9.94688    1.03023
## sj_dem_distancing_pre_post$`% over 125,000`                0.21367    0.02410
## sj_dem_distancing_pre_post$`% non hispanic/latino`         0.06225    0.02541
## sj_dem_distancing_pre_post$`percent associates or higher`  0.06934    0.03205
##                                                           t value Pr(>|t|)    
## (Intercept)                                                 9.655   <2e-16 ***
## sj_dem_distancing_pre_post$`% over 125,000`                 8.864   <2e-16 ***
## sj_dem_distancing_pre_post$`% non hispanic/latino`          2.450   0.0146 *  
## sj_dem_distancing_pre_post$`percent associates or higher`   2.164   0.0309 *  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 7.675 on 565 degrees of freedom
## Multiple R-squared:  0.3963, Adjusted R-squared:  0.3931 
## F-statistic: 123.6 on 3 and 565 DF,  p-value: < 2.2e-16

frac_model_inc_hisp_educ <- lm(sj_dem_distancing_pre_post$frac_increase ~ sj_dem_distancing_pre_post$`% over 125,000` +  sj_dem_distancing_pre_post$`% non hispanic/latino` + sj_dem_distancing_pre_post$`percent associates or higher`)
summary(frac_model_inc_hisp_educ)

## 
## Call:
## lm(formula = sj_dem_distancing_pre_post$frac_increase ~ sj_dem_distancing_pre_post$`% over 125,000` + 
##     sj_dem_distancing_pre_post$`% non hispanic/latino` + sj_dem_distancing_pre_post$`percent associates or higher`)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -1.95568 -0.29408  0.00776  0.26293  2.42973 
## 
## Coefficients:
##                                                           Estimate Std. Error
## (Intercept)                                               0.148617   0.070408
## sj_dem_distancing_pre_post$`% over 125,000`               0.015536   0.001647
## sj_dem_distancing_pre_post$`% non hispanic/latino`        0.003476   0.001736
## sj_dem_distancing_pre_post$`percent associates or higher` 0.005366   0.002190
##                                                           t value Pr(>|t|)    
## (Intercept)                                                 2.111   0.0352 *  
## sj_dem_distancing_pre_post$`% over 125,000`                 9.430   <2e-16 ***
## sj_dem_distancing_pre_post$`% non hispanic/latino`          2.002   0.0458 *  
## sj_dem_distancing_pre_post$`percent associates or higher`   2.450   0.0146 *  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.5245 on 565 degrees of freedom
## Multiple R-squared:  0.411,  Adjusted R-squared:  0.4079 
## F-statistic: 131.4 on 3 and 565 DF,  p-value: < 2.2e-16

When including education, percentage of Hispanic/Latino residents loses as much of its predictive power.

Multiple regression analysis: income, education, and white residents

difs_model_inc_white_educ <- lm(sj_dem_distancing_pre_post$`% increase in staying completely home` ~ sj_dem_distancing_pre_post$`% over 125,000` +  sj_dem_distancing_pre_post$`% white` + sj_dem_distancing_pre_post$`percent associates or higher`)
summary(difs_model_inc_white_educ)

## 
## Call:
## lm(formula = sj_dem_distancing_pre_post$`% increase in staying completely home` ~ 
##     sj_dem_distancing_pre_post$`% over 125,000` + sj_dem_distancing_pre_post$`% white` + 
##         sj_dem_distancing_pre_post$`percent associates or higher`)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -36.485  -3.912   1.201   4.756  17.949 
## 
## Coefficients:
##                                                           Estimate Std. Error
## (Intercept)                                               12.84118    0.94702
## sj_dem_distancing_pre_post$`% over 125,000`                0.23025    0.02370
## sj_dem_distancing_pre_post$`% white`                      -0.06215    0.01559
## sj_dem_distancing_pre_post$`percent associates or higher`  0.14009    0.02367
##                                                           t value Pr(>|t|)    
## (Intercept)                                                13.560  < 2e-16 ***
## sj_dem_distancing_pre_post$`% over 125,000`                 9.715  < 2e-16 ***
## sj_dem_distancing_pre_post$`% white`                       -3.986 7.60e-05 ***
## sj_dem_distancing_pre_post$`percent associates or higher`   5.918 5.64e-09 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 7.609 on 565 degrees of freedom
## Multiple R-squared:  0.4066, Adjusted R-squared:  0.4034 
## F-statistic:   129 on 3 and 565 DF,  p-value: < 2.2e-16

frac_model_inc_white_educ <- lm(sj_dem_distancing_pre_post$frac_increase ~ sj_dem_distancing_pre_post$`% over 125,000` +  sj_dem_distancing_pre_post$`% white` + sj_dem_distancing_pre_post$`percent associates or higher`)
summary(frac_model_inc_white_educ)

## 
## Call:
## lm(formula = sj_dem_distancing_pre_post$frac_increase ~ sj_dem_distancing_pre_post$`% over 125,000` + 
##     sj_dem_distancing_pre_post$`% white` + sj_dem_distancing_pre_post$`percent associates or higher`)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -1.93848 -0.29364  0.00696  0.26663  2.44543 
## 
## Coefficients:
##                                                            Estimate Std. Error
## (Intercept)                                               0.2204243  0.0655094
## sj_dem_distancing_pre_post$`% over 125,000`               0.0160183  0.0016395
## sj_dem_distancing_pre_post$`% white`                      0.0001366  0.0010786
## sj_dem_distancing_pre_post$`percent associates or higher` 0.0083123  0.0016373
##                                                           t value Pr(>|t|)    
## (Intercept)                                                 3.365 0.000818 ***
## sj_dem_distancing_pre_post$`% over 125,000`                 9.770  < 2e-16 ***
## sj_dem_distancing_pre_post$`% white`                        0.127 0.899277    
## sj_dem_distancing_pre_post$`percent associates or higher`   5.077 5.22e-07 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.5264 on 565 degrees of freedom
## Multiple R-squared:  0.4068, Adjusted R-squared:  0.4037 
## F-statistic: 129.2 on 3 and 565 DF,  p-value: < 2.2e-16

Multiple regression analysis: income, education, and Asian residents

difs_model_inc_asian_educ <- lm(sj_dem_distancing_pre_post$`% increase in staying completely home` ~ sj_dem_distancing_pre_post$`% over 125,000` +  sj_dem_distancing_pre_post$`% Asian` + sj_dem_distancing_pre_post$`percent associates or higher`)
summary(difs_model_inc_asian_educ)

## 
## Call:
## lm(formula = sj_dem_distancing_pre_post$`% increase in staying completely home` ~ 
##     sj_dem_distancing_pre_post$`% over 125,000` + sj_dem_distancing_pre_post$`% Asian` + 
##         sj_dem_distancing_pre_post$`percent associates or higher`)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -36.827  -3.820   1.117   4.945  18.232 
## 
## Coefficients:
##                                                           Estimate Std. Error
## (Intercept)                                                9.89236    0.90531
## sj_dem_distancing_pre_post$`% over 125,000`                0.21941    0.02348
## sj_dem_distancing_pre_post$`% Asian`                       0.07363    0.01515
## sj_dem_distancing_pre_post$`percent associates or higher`  0.10464    0.02342
##                                                           t value Pr(>|t|)    
## (Intercept)                                                10.927  < 2e-16 ***
## sj_dem_distancing_pre_post$`% over 125,000`                 9.346  < 2e-16 ***
## sj_dem_distancing_pre_post$`% Asian`                        4.859 1.53e-06 ***
## sj_dem_distancing_pre_post$`percent associates or higher`   4.469 9.52e-06 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 7.559 on 565 degrees of freedom
## Multiple R-squared:  0.4144, Adjusted R-squared:  0.4113 
## F-statistic: 133.3 on 3 and 565 DF,  p-value: < 2.2e-16

frac_model_inc_asian_educ <- lm(sj_dem_distancing_pre_post$frac_increase ~ sj_dem_distancing_pre_post$`% over 125,000` +  sj_dem_distancing_pre_post$`% Asian` + sj_dem_distancing_pre_post$`percent associates or higher`)
summary(frac_model_inc_asian_educ)

## 
## Call:
## lm(formula = sj_dem_distancing_pre_post$frac_increase ~ sj_dem_distancing_pre_post$`% over 125,000` + 
##     sj_dem_distancing_pre_post$`% Asian` + sj_dem_distancing_pre_post$`percent associates or higher`)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -1.9226 -0.2970  0.0086  0.2624  2.4582 
## 
## Coefficients:
##                                                            Estimate Std. Error
## (Intercept)                                               0.2158062  0.0630309
## sj_dem_distancing_pre_post$`% over 125,000`               0.0160168  0.0016345
## sj_dem_distancing_pre_post$`% Asian`                      0.0004213  0.0010550
## sj_dem_distancing_pre_post$`percent associates or higher` 0.0082464  0.0016303
##                                                           t value Pr(>|t|)    
## (Intercept)                                                 3.424 0.000662 ***
## sj_dem_distancing_pre_post$`% over 125,000`                 9.799  < 2e-16 ***
## sj_dem_distancing_pre_post$`% Asian`                        0.399 0.689844    
## sj_dem_distancing_pre_post$`percent associates or higher`   5.058 5.73e-07 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.5263 on 565 degrees of freedom
## Multiple R-squared:  0.407,  Adjusted R-squared:  0.4038 
## F-statistic: 129.3 on 3 and 565 DF,  p-value: < 2.2e-16

Multiple regression analysis: income, education, Asian residents, and English language ability

difs_model_inc_asian_educ_eng <- lm(sj_dem_distancing_pre_post$`% increase in staying completely home` ~ sj_dem_distancing_pre_post$`% over 125,000` +  sj_dem_distancing_pre_post$`% Asian` + sj_dem_distancing_pre_post$`percent associates or higher` + sj_dem_distancing_pre_post$`% speaking english > well`)
summary(difs_model_inc_asian_educ_eng)

## 
## Call:
## lm(formula = sj_dem_distancing_pre_post$`% increase in staying completely home` ~ 
##     sj_dem_distancing_pre_post$`% over 125,000` + sj_dem_distancing_pre_post$`% Asian` + 
##         sj_dem_distancing_pre_post$`percent associates or higher` + 
##         sj_dem_distancing_pre_post$`% speaking english > well`)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -37.412  -3.830   1.057   4.842  18.327 
## 
## Coefficients:
##                                                           Estimate Std. Error
## (Intercept)                                               14.10713    4.82041
## sj_dem_distancing_pre_post$`% over 125,000`                0.22493    0.02429
## sj_dem_distancing_pre_post$`% Asian`                       0.06323    0.01914
## sj_dem_distancing_pre_post$`percent associates or higher`  0.12258    0.03090
## sj_dem_distancing_pre_post$`% speaking english > well`    -0.05580    0.06269
##                                                           t value Pr(>|t|)    
## (Intercept)                                                 2.927  0.00357 ** 
## sj_dem_distancing_pre_post$`% over 125,000`                 9.261  < 2e-16 ***
## sj_dem_distancing_pre_post$`% Asian`                        3.304  0.00101 ** 
## sj_dem_distancing_pre_post$`percent associates or higher`   3.967 8.23e-05 ***
## sj_dem_distancing_pre_post$`% speaking english > well`     -0.890  0.37373    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 7.561 on 564 degrees of freedom
## Multiple R-squared:  0.4152, Adjusted R-squared:  0.411 
## F-statistic: 100.1 on 4 and 564 DF,  p-value: < 2.2e-16

frac_model_inc_asian_educ_eng <- lm(sj_dem_distancing_pre_post$frac_increase ~ sj_dem_distancing_pre_post$`% over 125,000` +  sj_dem_distancing_pre_post$`% Asian` + sj_dem_distancing_pre_post$`percent associates or higher` + sj_dem_distancing_pre_post$`% speaking english > well`)
summary(frac_model_inc_asian_educ_eng)

## 
## Call:
## lm(formula = sj_dem_distancing_pre_post$frac_increase ~ sj_dem_distancing_pre_post$`% over 125,000` + 
##     sj_dem_distancing_pre_post$`% Asian` + sj_dem_distancing_pre_post$`percent associates or higher` + 
##     sj_dem_distancing_pre_post$`% speaking english > well`)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -1.95085 -0.29845  0.01225  0.26295  2.42410 
## 
## Coefficients:
##                                                             Estimate Std. Error
## (Intercept)                                                0.5717896  0.3355027
## sj_dem_distancing_pre_post$`% over 125,000`                0.0164831  0.0016904
## sj_dem_distancing_pre_post$`% Asian`                      -0.0004575  0.0013321
## sj_dem_distancing_pre_post$`percent associates or higher`  0.0097624  0.0021509
## sj_dem_distancing_pre_post$`% speaking english > well`    -0.0047132  0.0043629
##                                                           t value Pr(>|t|)    
## (Intercept)                                                 1.704   0.0889 .  
## sj_dem_distancing_pre_post$`% over 125,000`                 9.751  < 2e-16 ***
## sj_dem_distancing_pre_post$`% Asian`                       -0.343   0.7314    
## sj_dem_distancing_pre_post$`percent associates or higher`   4.539 6.92e-06 ***
## sj_dem_distancing_pre_post$`% speaking english > well`     -1.080   0.2805    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.5262 on 564 degrees of freedom
## Multiple R-squared:  0.4082, Adjusted R-squared:  0.404 
## F-statistic: 97.27 on 4 and 564 DF,  p-value: < 2.2e-16

Conclusion from multiple regression analyses on change in leaving home behavior

From the results presented above, we see that income (making over $125,000) predicts about 36% of the variability in percent of devices leaving the home across blockgroups. Adding in education leads to a prediction of about 39% of the variation, and including percent of residents that are Asian with both education and income adds about 2% predictive power. Additional variables on top of these three negligibly increase the percent of variation in the data that the variables explain.

Testing animating the plot

# another collection for pre shelter in place behavior
sj_dem_distancing_pre_shelter <- sj_dem_distancing %>% 
  dplyr::select(-`% not completely at home`) %>%
  left_join(sj_internet_by_block %>% dplyr::select(`% not completely at home pre shelter`, blockgroup))

# relabel column for leaving home
colnames(sj_dem_distancing_pre_shelter)[18] <- "% not completely at home"

sj_dem_distancing[is.na(sj_dem_distancing)] <- 0
sj_dem_distancing_pre_shelter[is.na(sj_dem_distancing_pre_shelter)] <- 0

# add column indicating before or after shelter in place, then bind the two sets of data
sj_dem_distancing_pre_shelter <- sj_dem_distancing_pre_shelter %>% 
  mutate(
    income_trendline =
      fitted(lm(sj_dem_distancing_pre_shelter$`% not completely at home` ~ sj_dem_distancing_pre_shelter$`% over 125,000`))
  ) %>% 
  cbind(`Before or After Shelter-in-Place` = "before")
sj_dem_distancing <-
  sj_dem_distancing %>%
  mutate(
    income_trendline =
      fitted(lm(sj_dem_distancing$`% not completely at home` ~ sj_dem_distancing$`% over 125,000`))
  ) %>% 
  cbind(`Before or After Shelter-in-Place` = "after") %>% 
  rbind(sj_dem_distancing_pre_shelter)

# try animating
fig <- 
  plot_ly (sj_dem_distancing) %>%
    add_trace(
      x = ~`% over 125,000`, 
      y = ~`% not completely at home`, 
      frame = ~`Before or After Shelter-in-Place`, 
      type = 'scatter', 
      mode = 'markers'
    ) %>% 
    add_trace(
      x = ~`% over 125,000`,
      y = ~income_trendline,
      type = 'scatter',
      mode = 'lines',
      line = list(size = 5, color = 'rgba(255, 165, 0, 1)'),
      frame = ~`Before or After Shelter-in-Place`
    ) %>% 
  animation_button(visible = F)
fig

# # save as rds
# saveRDS(sj_dem_distancing, "/Users/simonespeizer/pCloud Drive/Shared/SFBI/Restricted Data Library/Safegraph/covid19analysis/sj_sd_dem_data.rds")


# fig <- plot_ly(sj_dem_distancing) %>% 
#   add_trace(
#     x = ~`% over 125,000`,
#     y = ~`% not completely at home`,
#     frame = ~`Before or After Shelter-in-Place`,
#     type = "scatter",
#     mode = "markers",
#     name = "Under $125,000",
#     marker = list(size = 5, color = 'rgba(50, 150, 200, 1)'),
#     visible = T
#   ) %>% 
#   add_trace(
#     x = ~`% over 125,000`,
#     y = fitted(lm(sj_dem_distancing$`% not completely at home` ~ sj_dem_distancing$`% over 125,000`)),
#     name = 'trendline',
#     mode = 'lines',
#     line = list(size = 5, color = 'rgba(255, 165, 0, 1)'),
#     frame = ~`Before or After Shelter-in-Place`,
#     visible = F
#   ) %>%
#   add_trace(
#     x = ~`% not speaking spanish`,
#     y = ~`% not completely at home`,
#     frame = ~`Before or After Shelter-in-Place`,
#     name = "speak spanish",
#     marker = list(size = 5, color = 'rgba(50, 150, 200, 1)'),
#     visible = F
#   ) %>% 
#   add_trace(
#     x = ~`% not speaking spanish`,
#     y = fitted(lm(sj_dem_distancing$`% not completely at home` ~ sj_dem_distancing$`% not speaking spanish`)),
#     name = 'trendline',
#     mode = 'lines',
#     line = list(size = 5, color = 'rgba(255, 165, 0, 1)'),
#     frame = ~`Before or After Shelter-in-Place`,
#     visible = F
#   ) %>% 
#   add_trace(
#     x = ~`percent associates or higher`,
#     y = ~`% not completely at home`,
#     frame = ~`Before or After Shelter-in-Place`,
#     name = "percent higher degree",
#     marker = list(size = 5, color = 'rgba(50, 150, 200, 1)'),
#     visible = F
#   ) %>% 
#   add_trace(
#     x = ~`percent associates or higher`,
#     y = fitted(lm(sj_dem_distancing$`% not completely at home` ~ sj_dem_distancing$`percent associates or higher`)),
#     name = 'trendline',
#     mode = 'lines',
#     line = list(size = 5, color = 'rgba(255, 165, 0, 1)'),
#     frame = ~`Before or After Shelter-in-Place`,
#     visible = F
#   ) %>%
#   add_trace(
#     x = ~`percent high speed`,
#     y = ~`% not completely at home`,
#     frame = ~`Before or After Shelter-in-Place`,
#     name = "percent high speed internet access",
#     marker = list(size = 5, color = 'rgba(50, 150, 200, 1)'),
#     visible = F
#   ) %>% 
#   add_trace(
#     x = ~`percent high speed`,
#     y = fitted(lm(sj_dem_distancing$`% not completely at home` ~ sj_dem_distancing$`percent high speed`)),
#     name = 'trendline',
#     mode = 'lines',
#     line = list(size = 5, color = 'rgba(255, 165, 0, 1)'),
#     frame = ~sj_dem_distancing$`Before or After Shelter-in-Place`,
#     visible = F
#   ) %>%
#   add_trace(
#     x = ~`percent less than 30`,
#     y = ~`% not completely at home`,
#     frame = ~`Before or After Shelter-in-Place`,
#     name = "percent less than 30",
#     marker = list(size = 5, color = 'rgba(50, 150, 200, 1)'),
#     visible = F
#   ) %>% 
#   add_trace(
#     x = ~`percent less than 30`,
#     y = fitted(lm(sj_dem_distancing$`% not completely at home` ~ sj_dem_distancing$`percent less than 30`)),
#     name = 'trendline',
#     mode = 'lines',
#     line = list(size = 5, color = 'rgba(255, 165, 0, 1)'),
#     frame = ~`Before or After Shelter-in-Place`,
#     visible = F
#   ) %>%
#   layout(
#     updatemenus = list(
#       list(
#         active = 0,
#         type = 'buttons',
#         buttons = list(
#           list(
#             label = "Households Under $125,000",
#             method = 'update',
#             args = list(list(visible = c(T, T, F, F, F, F, F, F, F, F)),
#                         list(title = "Under $125,000",
#                              xaxis = list(title = "% Households Under $125,000 in Income")))),
#           list(
#             label = "Speaking Spanish",
#             method = 'update',
#             args = list(list(visible = c(F, F, T, T, F, F, F, F, F, F)),
#                         list(title = "Not Speaking Spanish",
#                              xaxis = list(title = "% Residents Not Speaking Spanish")))),
#           list(
#             label = "Education Level",
#             method = 'update',
#             args= list(list(visible = c(F, F, F, F, T, T, F, F, F, F)),
#                        list(xaxis = list(title = "% Residents With Associate's Degree or Higher")))),
#           list(
#             label = "High Speed Internet",
#             method = 'update',
#             args= list(list(visible = c(F, F, F, F, F, F, T, T, F, F)),
#                        list(xaxis = list(title = "% Households With High Speed Internet Access")))),
#           list(
#             label = "Young Population",
#             method = 'update',
#             args= list(list(visible = c(F, F, F, F, F, F, T, T, F, F)),
#                        list(xaxis = list(title = "% Residents Under Age 30"))))
#           )
#           )
#         ),
#     yaxis = list(title = "% Residents Leaving Home", 
#                  font = list(size = 15)),
#     showlegend = FALSE
#       )
# fig

Experimentation

Experimentation with other variables and other ways of analyzing the social distancing data. First I look at a few other possible variables. I start with units in the structure.

# try getting other variables
# get data on units in structure
sj_units_in_structure_by_block <- getCensus(
    name = "acs/acs5",
    vintage = 2018,
    region = "block group:*", 
    regionin = "state:06+county:085",
    vars = "group(B25024)"
  ) %>% 
  mutate(
    blockgroup =
      paste0(state,county,tract,block_group)
  ) %>% 
  select_if(!names(.) %in% c("GEO_ID","state","county","tract","block_group","NAME")) %>% 
  select(-c(contains("EA"),contains("MA"),contains("M"))) %>% 
  gather(key = "variable", value = "estimate", -blockgroup) %>% 
  mutate(label = acs_vars$label[match(variable,acs_vars$name)]) %>% 
  select(-variable) %>% 
  separate(label, into = c(NA, NA, "units"), sep = "!!") %>% 
  filter(!is.na(units)) %>%
  spread(key = units, value = estimate) %>%
  mutate(total_nums = `1, attached` + `1, detached` + `10 to 19` + `2` + `20 to 49`+ `3 or 4` + `5 to 9`+ `50 or more`+ `Boat, RV, van, etc.`+ `Mobile home`, `percent 20 or more` = (`20 to 49`+`50 or more`)* 100/ total_nums, `percent 1 only` = (`1, attached` + `1, detached`)*100/total_nums, `percent > 1` = 100 - `percent 1 only`) %>%
  left_join(sj_age_by_block %>% dplyr::select_if(!names(.) %in% c("elderly", "percent elderly", "less than 30", "percent less than 30"))) %>%
  filter(!is.na(device_count))

# plot 
sj_units_in_structure_by_block %>% 
  ggplot(aes(
  x = `percent 20 or more`,
  y = `% not completely at home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of structures with more than 20 units",
    y = "Percent devices leaving home on weekdays since shelter-in-place",
    title = "San Jose: Social Distancing and 20 or More Units Per Structure"
  )

summary(lm(`% not completely at home` ~ `percent 20 or more`, sj_units_in_structure_by_block))

## 
## Call:
## lm(formula = `% not completely at home` ~ `percent 20 or more`, 
##     data = sj_units_in_structure_by_block)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -26.551  -5.052  -0.451   4.274  37.149 
## 
## Coefficients:
##                      Estimate Std. Error t value Pr(>|t|)    
## (Intercept)          51.15091    0.40431 126.513   <2e-16 ***
## `percent 20 or more`  0.01873    0.02026   0.924    0.356    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 8.354 on 566 degrees of freedom
##   (1 observation deleted due to missingness)
## Multiple R-squared:  0.001507,   Adjusted R-squared:  -0.0002566 
## F-statistic: 0.8545 on 1 and 566 DF,  p-value: 0.3557

sj_units_in_structure_by_block %>% 
  ggplot(aes(
  x = `percent 1 only`,
  y = `% not completely at home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of structures with only one unit",
    y = "Percent devices leaving home on weekdays since shelter-in-place",
    title = "San Jose: Social Distancing and Only 1 Unit Per Structure"
  )

summary(lm(`% not completely at home` ~ `percent 1 only`, sj_units_in_structure_by_block))

## 
## Call:
## lm(formula = `% not completely at home` ~ `percent 1 only`, data = sj_units_in_structure_by_block)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -25.452  -5.117  -0.322   4.364  38.053 
## 
## Coefficients:
##                  Estimate Std. Error t value Pr(>|t|)    
## (Intercept)      55.38663    0.88348  62.692  < 2e-16 ***
## `percent 1 only` -0.05596    0.01125  -4.975 8.68e-07 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 8.184 on 566 degrees of freedom
##   (1 observation deleted due to missingness)
## Multiple R-squared:  0.04189,    Adjusted R-squared:  0.0402 
## F-statistic: 24.75 on 1 and 566 DF,  p-value: 8.681e-07

Household type and size:

# load data on household type and size
sj_house_size_type_by_block <- getCensus(
    name = "acs/acs5",
    vintage = 2018,
    region = "block group:*", 
    regionin = "state:06+county:085",
    vars = "group(B11016)"
  ) %>% 
  mutate(
    blockgroup =
      paste0(state,county,tract,block_group)
  ) %>% 
  select_if(!names(.) %in% c("GEO_ID","state","county","tract","block_group","NAME")) %>% 
  select(-c(contains("EA"),contains("MA"),contains("M"))) %>% 
  gather(key = "variable", value = "estimate", -blockgroup) %>% 
  mutate(label = acs_vars$label[match(variable,acs_vars$name)]) %>% 
  select(-variable) %>% 
  separate(label, into = c(NA, NA, "type", "size"), sep = "!!") %>% 
  filter(!is.na(type))


# household type
sj_house_type_by_block <- sj_house_size_type_by_block %>% 
  filter(is.na(size)) %>% 
  dplyr::select(-size) %>%
  spread(key = type, value = estimate) %>% 
  mutate(`total households` = `Family households` + `Nonfamily households`, `percent nonfamily` = `Nonfamily households` / `total households`) %>%
  left_join(sj_age_by_block %>% dplyr::select_if(!names(.) %in% c("elderly", "percent elderly", "less than 30", "percent less than 30"))) %>%
  filter(!is.na(device_count))

sj_house_type_by_block %>% 
  ggplot(aes(
  x = `percent nonfamily`,
  y = `% not completely at home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent nonfamily households",
    y = "Percent devices leaving home on weekdays since shelter-in-place",
    title = "San Jose: Social Distancing and Household Type"
  )

summary(lm(`% not completely at home` ~ `percent nonfamily`, sj_house_type_by_block))

## 
## Call:
## lm(formula = `% not completely at home` ~ `percent nonfamily`, 
##     data = sj_house_type_by_block)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -24.616  -5.108  -0.157   4.261  39.714 
## 
## Coefficients:
##                     Estimate Std. Error t value Pr(>|t|)    
## (Intercept)          48.8864     0.6327  77.262  < 2e-16 ***
## `percent nonfamily`  10.1400     2.1962   4.617 4.82e-06 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 8.208 on 566 degrees of freedom
##   (1 observation deleted due to missingness)
## Multiple R-squared:  0.0363, Adjusted R-squared:  0.03459 
## F-statistic: 21.32 on 1 and 566 DF,  p-value: 4.82e-06

# household size
sj_house_size_by_block <- sj_house_size_type_by_block %>% 
  filter(!is.na(size)) %>% 
  dplyr::select(-type) %>%
  group_by(blockgroup, size) %>%
  summarize(`total of this size` = sum(estimate)) %>% 
  spread(key = size, value = `total of this size`) %>%
  mutate(total_nums = `1-person household` + `2-person household` + `3-person household` + `4-person household` + `5-person household`+ `6-person household` + `7-or-more person household`, `percent 5 or more` = (`5-person household`+`6-person household` + `7-or-more person household`)* 100/ total_nums, `percent 1 or 2 only` = (`1-person household` + `2-person household`)*100/total_nums) %>%
  left_join(sj_age_by_block %>% dplyr::select_if(!names(.) %in% c("elderly", "percent elderly", "less than 30", "percent less than 30"))) %>%
  filter(!is.na(device_count))

sj_house_size_by_block %>% 
  ggplot(aes(
  x = `percent 5 or more`,
  y = `% not completely at home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of households with 5 or more people",
    y = "Percent devices leaving home on weekdays since shelter-in-place",
    title = "San Jose: Social Distancing and Households With 5 or More"
  )

summary(lm(`% not completely at home` ~ `percent 5 or more`, sj_house_size_by_block))

## 
## Call:
## lm(formula = `% not completely at home` ~ `percent 5 or more`, 
##     data = sj_house_size_by_block)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -26.136  -4.979  -0.518   4.175  36.154 
## 
## Coefficients:
##                     Estimate Std. Error t value Pr(>|t|)    
## (Intercept)         49.85951    0.56004  89.028  < 2e-16 ***
## `percent 5 or more`  0.08454    0.02513   3.364 0.000821 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 8.278 on 566 degrees of freedom
##   (1 observation deleted due to missingness)
## Multiple R-squared:  0.0196, Adjusted R-squared:  0.01786 
## F-statistic: 11.31 on 1 and 566 DF,  p-value: 0.0008215

sj_house_size_by_block %>% 
  ggplot(aes(
  x = `percent 1 or 2 only`,
  y = `% not completely at home`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of households with 1 or 2 people",
    y = "Percent devices leaving home on weekdays since shelter-in-place",
    title = "San Jose: Social Distancing and Small Household Size"
  )

summary(lm(`% not completely at home` ~ `percent 1 or 2 only`, sj_house_size_by_block))

## 
## Call:
## lm(formula = `% not completely at home` ~ `percent 1 or 2 only`, 
##     data = sj_house_size_by_block)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -25.847  -5.127  -0.450   4.596  37.830 
## 
## Coefficients:
##                       Estimate Std. Error t value Pr(>|t|)    
## (Intercept)           50.13694    0.96771   51.81   <2e-16 ***
## `percent 1 or 2 only`  0.02678    0.02013    1.33    0.184    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 8.348 on 566 degrees of freedom
##   (1 observation deleted due to missingness)
## Multiple R-squared:  0.003118,   Adjusted R-squared:  0.001356 
## F-statistic:  1.77 on 1 and 566 DF,  p-value: 0.1839

Next I consider different ways of looking at the social distancing data. First I try distance traveled.

# try other ways of looking at the social distancing data
# first look at total distance traveled
sj_sd_distance <- sj_socialdistancing %>% 
  filter(date > shelter_start) %>% 
  group_by(origin_census_block_group) %>% 
  summarize(total_dist_traveled = sum(distance_traveled_from_home), device_count = sum(device_count)) %>%
  mutate(total_dist_per_device = total_dist_traveled / device_count)

sj_distance_testing <- left_join(sj_ami_by_block, sj_sd_distance, by = c("blockgroup" = "origin_census_block_group")) %>% left_join(sj_age_by_block %>% select(blockgroup, `percent less than 30`))

sj_distance_testing %>% filter(total_dist_per_device < 500)  %>% 
  ggplot(aes(
  x = `% over 75,000`,
  y = total_dist_per_device
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of housholds with incomes over $75,000 (50% AMI) annually",
    y = "Average distance traveled per device during weekdays since shelter-in-place",
    title = "San Jose: Social Distancing and Income, Distance Metric"
  )

This is very skewed by outliers, and probably not a useful metric.

Now I consider including devices that traveled <1km as staying at (or near) home.

sj_sd_range <- sj_socialdistancing %>% 
  filter(weekend == F) %>% 
  filter(date > shelter_start) %>%
  mutate(travel_buckets_split = lapply(bucketed_distance_traveled, function(x) strsplit(x, "<1000")[[1]][2]), less_than_1km = lapply(travel_buckets_split, function(x) strsplit(x, ":")[[1]][2]), less_than_1km = lapply(less_than_1km, function(x) strsplit(x, ",")[[1]][1])) %>%
  mutate(less_than_1km = lapply(less_than_1km, function(x) str_remove(x, "[}]")))  %>% # clean a bit more
  mutate(less_than_1km = as.numeric(less_than_1km), less_than_1km = replace_na(less_than_1km, 0)) %>% 
  mutate(home_or_1km = completely_home_device_count + less_than_1km) %>% 
  group_by(origin_census_block_group) %>% 
  summarize(home_or_1km = sum(home_or_1km), device_count = sum(device_count)) %>% 
  mutate(`% Within 1km of Home` = (home_or_1km/device_count*100) %>% round(1), `% farther than 1km` = (100-`% Within 1km of Home`))

# join this with other data
sj_1km_testing <- left_join(sj_ami_by_block, sj_sd_range, by = c("blockgroup" = "origin_census_block_group")) %>% 
  left_join(sj_occupants_per_room_by_block %>% dplyr::select(`percent less than 1`, blockgroup)) %>%
  left_join(sj_age_by_block %>% dplyr::select(`percent less than 30`, blockgroup)) %>%
  left_join(sj_lang_by_block %>% dplyr::select(`% speaking english > well`, blockgroup)) 

# plot with income
sj_1km_testing %>%  
  ggplot(aes(
  x = `% over 75,000`,
  y = `% farther than 1km`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of housholds with incomes over $75,000 (50% AMI) annually",
    y = "% of devices going farther than 1km of home, weekdays since shelter-in-place",
    title = "San Jose: Social Distancing and Income, 1km Range"
  )

summary(lm(`% farther than 1km` ~ `% over 75,000`, sj_1km_testing))

## 
## Call:
## lm(formula = `% farther than 1km` ~ `% over 75,000`, data = sj_1km_testing)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -21.389  -4.795  -0.606   4.185  34.925 
## 
## Coefficients:
##                 Estimate Std. Error t value Pr(>|t|)    
## (Intercept)     64.35928    1.11518   57.71   <2e-16 ***
## `% over 75,000` -0.20909    0.01719  -12.16   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 7.444 on 566 degrees of freedom
##   (1 observation deleted due to missingness)
## Multiple R-squared:  0.2072, Adjusted R-squared:  0.2058 
## F-statistic:   148 on 1 and 566 DF,  p-value: < 2.2e-16

# plot with age
sj_1km_testing %>%  
  ggplot(aes(
  x = `percent less than 30`,
  y = `% farther than 1km`
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of people younger than 30",
    y = "Percent of devices farther than 1km of home during weekdays since shelter-in-place",
    title = "San Jose: Social Distancing and Age, 1km Range"
  )

summary(lm(`% farther than 1km` ~ `percent less than 30`, sj_1km_testing))

## 
## Call:
## lm(formula = `% farther than 1km` ~ `percent less than 30`, data = sj_1km_testing)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -28.505  -4.978  -0.362   4.274  39.808 
## 
## Coefficients:
##                        Estimate Std. Error t value Pr(>|t|)    
## (Intercept)            44.49223    1.51348  29.397  < 2e-16 ***
## `percent less than 30`  0.17973    0.03836   4.685 3.51e-06 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 8.299 on 567 degrees of freedom
## Multiple R-squared:  0.03726,    Adjusted R-squared:  0.03557 
## F-statistic: 21.95 on 1 and 567 DF,  p-value: 3.513e-06

# run multiple regression model
modeltest2 <- lm(sj_1km_testing$`% farther than 1km` ~ sj_1km_testing$`% over 75,000` + sj_1km_testing$`percent less than 30` + sj_1km_testing$`% speaking english > well` + sj_1km_testing$`percent less than 1`)
summary(modeltest2)

## 
## Call:
## lm(formula = sj_1km_testing$`% farther than 1km` ~ sj_1km_testing$`% over 75,000` + 
##     sj_1km_testing$`percent less than 30` + sj_1km_testing$`% speaking english > well` + 
##     sj_1km_testing$`percent less than 1`)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -20.646  -4.675  -0.762   4.296  34.809 
## 
## Coefficients:
##                                            Estimate Std. Error t value Pr(>|t|)
## (Intercept)                                58.57603    4.69256  12.483   <2e-16
## sj_1km_testing$`% over 75,000`             -0.21745    0.02166 -10.039   <2e-16
## sj_1km_testing$`percent less than 30`       0.03813    0.04267   0.894   0.3718
## sj_1km_testing$`% speaking english > well`  0.09468    0.04607   2.055   0.0403
## sj_1km_testing$`percent less than 1`       -0.03946    0.04679  -0.843   0.3994
##                                               
## (Intercept)                                ***
## sj_1km_testing$`% over 75,000`             ***
## sj_1km_testing$`percent less than 30`         
## sj_1km_testing$`% speaking english > well` *  
## sj_1km_testing$`percent less than 1`          
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 7.424 on 563 degrees of freedom
##   (1 observation deleted due to missingness)
## Multiple R-squared:  0.2158, Adjusted R-squared:  0.2102 
## F-statistic: 38.72 on 4 and 563 DF,  p-value: < 2.2e-16

It looks like the fit of these selected variables is slightly better for the social distancing data based on not traveling farther than 1km.

Now I also consider “non-work” behavior.

sj_nonworking_by_block <- sj_socialdistancing %>% 
  filter(weekend == F) %>% 
  filter(date > shelter_start) %>%
  mutate(nonworking = device_count - completely_home_device_count - part_time_work_behavior_devices - full_time_work_behavior_devices) %>%
  group_by(origin_census_block_group) %>%
  summarize(nonworking_count = sum(nonworking), total_device = sum(device_count)) %>% 
  mutate(nonworking_percent = nonworking_count*100 / total_device, percent_only_work_home = 100-nonworking_percent) %>%
  left_join(sj_1km_testing %>% dplyr::select(`% over 75,000`, `percent less than 30`, `% speaking english > well`, `percent less than 1`, blockgroup), by = c("origin_census_block_group" = "blockgroup"))


# plot against age and income
sj_nonworking_by_block %>%  
  ggplot(aes(
  x = `% over 75,000`,
  y = nonworking_percent
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of housholds with incomes over $75,000 (50% AMI) annually",
    y = "Percent of devices leaving home for non-work purposes during weekdays since shelter-in-place",
    title = "San Jose: Social Distancing and Income, Nonworking Behavior"
  )

summary(lm(nonworking_percent ~ `% over 75,000`, sj_nonworking_by_block))

## 
## Call:
## lm(formula = nonworking_percent ~ `% over 75,000`, data = sj_nonworking_by_block)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -19.922  -4.227  -0.758   3.670  33.628 
## 
## Coefficients:
##                 Estimate Std. Error t value Pr(>|t|)    
## (Intercept)     52.99603    1.02994   51.46   <2e-16 ***
## `% over 75,000` -0.16521    0.01588  -10.41   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 6.875 on 566 degrees of freedom
##   (1 observation deleted due to missingness)
## Multiple R-squared:  0.1606, Adjusted R-squared:  0.1591 
## F-statistic: 108.3 on 1 and 566 DF,  p-value: < 2.2e-16

sj_nonworking_by_block %>%  
  ggplot(aes(
  x = `percent less than 30`,
  y = nonworking_percent
)) + geom_point() + 
  geom_smooth(method=lm) + 
  labs(
    x = "Percent of people younger than 30",
    y = "Percent of devices leaving home for non-work purposes during weekdays since shelter-in-place",
    title = "San Jose: Social Distancing and Age, Nonworking Behavior"
  )

summary(lm(nonworking_percent ~ `percent less than 30`, sj_nonworking_by_block))

## 
## Call:
## lm(formula = nonworking_percent ~ `percent less than 30`, data = sj_nonworking_by_block)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -22.773  -4.144  -0.345   3.400  32.456 
## 
## Coefficients:
##                        Estimate Std. Error t value Pr(>|t|)    
## (Intercept)            35.85635    1.33549  26.849  < 2e-16 ***
## `percent less than 30`  0.17865    0.03385   5.277 1.87e-07 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 7.323 on 567 degrees of freedom
## Multiple R-squared:  0.04682,    Adjusted R-squared:  0.04513 
## F-statistic: 27.85 on 1 and 567 DF,  p-value: 1.871e-07

# multiple regression model
modeltest3 <- lm(sj_nonworking_by_block$nonworking_percent ~ sj_nonworking_by_block$`% over 75,000` + sj_nonworking_by_block$`percent less than 30` + sj_nonworking_by_block$`% speaking english > well` + sj_nonworking_by_block$`percent less than 1`)
summary(modeltest3)

## 
## Call:
## lm(formula = sj_nonworking_by_block$nonworking_percent ~ sj_nonworking_by_block$`% over 75,000` + 
##     sj_nonworking_by_block$`percent less than 30` + sj_nonworking_by_block$`% speaking english > well` + 
##     sj_nonworking_by_block$`percent less than 1`)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -18.758  -4.001  -0.742   3.843  31.685 
## 
## Coefficients:
##                                                    Estimate Std. Error t value
## (Intercept)                                        44.10456    4.31944  10.211
## sj_nonworking_by_block$`% over 75,000`             -0.17101    0.01994  -8.577
## sj_nonworking_by_block$`percent less than 30`       0.08287    0.03928   2.110
## sj_nonworking_by_block$`% speaking english > well`  0.07965    0.04241   1.878
## sj_nonworking_by_block$`percent less than 1`       -0.01102    0.04307  -0.256
##                                                    Pr(>|t|)    
## (Intercept)                                          <2e-16 ***
## sj_nonworking_by_block$`% over 75,000`               <2e-16 ***
## sj_nonworking_by_block$`percent less than 30`        0.0353 *  
## sj_nonworking_by_block$`% speaking english > well`   0.0609 .  
## sj_nonworking_by_block$`percent less than 1`         0.7981    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 6.833 on 563 degrees of freedom
##   (1 observation deleted due to missingness)
## Multiple R-squared:  0.1752, Adjusted R-squared:  0.1693 
## F-statistic: 29.89 on 4 and 563 DF,  p-value: < 2.2e-16

These variables do worse for the percent nonworking metric, which makes sense.